close

Вход

Забыли?

вход по аккаунту

?

MICROWAVE IMAGE UNDERSTANDING AND ITS APPLICATION TO RADAR CROSS-SECTION MANAGEMENT

код для вставкиСкачать
INFORMATION TO USERS
The most advanced technology has been used to photo­
graph and reproduce this manuscript from the microfilm
master. UMI films the original text directly from the copy
submitted. Thus, some dissertation copies are in typewriter
face, while others may be from a computer printer.
In the unlikely event that the author did not send UMI a
complete manuscript and there are missing pages, these will
be noted. Also, if unauthorized copyrighted material had to
be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are re­
produced by sectioning the original, beginning at the upper
left-hand comer and continuing from left to right in equal
sections with small overlaps. Each oversize page is available
as one exposure on a standard 35 mm slide or as a 17" x 23"
black and white photographic print for an additional charge.
Photographs included in the original manuscript have been
reproduced xerographically in this copy. 35 mm slides or
6" x 9" black and white photographic prints are available for
any photographs or illustrations appearing in this copy for
an additional charge. Contact UMI directly to order.
UMI
A c c e s s in g th e World's Information s in c e 1938
3 0 0 North Z e eb Road, Ann Arbor, Ml 48 1 0 6 1346 USA
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Order N um ber 8725182
Microwave image understanding and its application to radar
cross section management
Li, Hsueh-Jyh, Ph.D.
University of Pennsylvania, 1987
UMI
300 N. Zeeb Rd.
Ann Arbor, MI 48106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
PLEASE NOTE:
In all cases this material has been filmed in the best possible way from the available copy.
Problems encountered with this document have been identified here with a check mark
1.
Glossy photographs or pages______
2.
Colored illustrations, paper or print_______
3.
Photographs with dark background_____
4.
Illustrations are poor copy
5.
Pages with black marks, not original copy
6.
Print shows through as there is text on both sides of p a g e_______
7.
Indistinct, broken or small print on severalpages___
8.
Print exceeds margin requirements______
9.
Tightly bound copy with print lost in spine_______
V
.
^
^
10.
Computer printout pages with indistinct print______
11.
Page(s)___________ lacking when material received, and not available from school or
author.
12.
Page(s)___________ seem to be missing in numbering only as text follows.
13.
Two pages numbered
14.
Curling and wrinkled pages______
15.
Dissertation contains pages with print at a slant, filmed a s received_________
16.
Other_______________________________________________________________________
. Text follows.
University
Microfilms
International
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
MICROWAVE IMAGE UNDERSTANDING
AND ITS APPLICATION TO
RADAR CROSS SECTION MANAGEMENT
Hsueh-Jyh Li
A DISSERTATION
in
Electrical Engineering
Presented to the Faculties of the University of Pennsylvania in
Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy.
1987
Supervisor of Dissertation
Graduate Group Chairperson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In remembrance o f
My parents and my brother Hsueh-Jei Li
-
ii -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknowledgements
There are so many people to whom I would like to express my gratitude for
making this dissertation possible. My deepest gratitude goes to my advisor, Professor
Nabil Farhat, who initiated my study in the realm of microwave imaging and other
related research areas and is the source for many of the ideas discussed in this disser­
tation. He has also provided the financial support in the course of the study. I thank
.my fellow graduate students in the Electro-Optics and Microwave Optics Laboratory,
in particular, Dr. Yuhsyen Shen, Dr. Charlie Werner, Zon-yin Shae, Baocheng Bai,
Ken Schultz, and Kang-Suk Lee, who have provided stimulating discussions and
incisive criticisms, and friendships as well, all of which made my study at Penn a
pleasant experience. I also wish to thank Prof. Jaggard for stimulating discussions.
Sitting in his class is really an enjoyment
I especially thank to my beloved wife, Su-Jean, without her patience, under­
standing, and encouragement, this work would not have become a reality. Finally, I
must express my apology for the many hours of neglect my son, Harry. He deserves
definitely more from me.
ui -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
MICROWAVE IMAGE UNDERSTANDING AND ITS APPLICATION TO
RADAR CROSS SECTION MANAGEMENT
Hsueh-Jyh Li
Nabil H. Farhat
In this dissertation the basic scattering properties of a perfectly conducting
object are briefly reviewed. In the high frequency region, the scattered field of a
complex shaped object can be attributed to a combination of several mechanisms.
The advent of high resolution radar enables us to consider each scattering mechanism
separately. The microwave image of a conducting object is interpreted from a new
approach, based on the understanding of the scattering mechanism and the image
reconstruction algorithm. The connection between various scattering mechanisms and
their reconstructed images is then established. From this we can interpret what the
image represents and predict what the image will look like for given spectral and
angular windows. Several numerical and experimental examples have been included
to support this new interpretation approach.
A new algorithm to extrapolate the available data into the missing bands is dev­
ised. Both simulation and experimental results have shown the effectiveness of this
method in microwave diversity radar imaging. A significant application of this new
algorithm is in identifying the hot spots of a target when the bandwidth of the imag­
ing system is too narrow' to give acceptable resolution. The satisfactory interpretation
of microwave images and the effectiveness of the extrapolation algorithm devised are
fundamental to the study of RCS management and target recognition.
A new term, "diaphanization", defined as the techniques of reducing RCS and
the techniques of obscuring an image is introduced. RCS management is considered
not only from a detection perspective, namely the reduction of the target’s RCS to
elude radar detection, but also from the image point of view where disguising of a
- iv -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
target’s appearance to impede recognition by an imaging radar is sought. The pro­
cedure which employs microwave diversity imaging to diaphanize a target over
prescribed spectral and angular windows is given and the robustness of the diversity
imaging system to the Gaussian noise is demonstrated. Some rules for distorting an
image are proposed. These rules are: create artificial discontinuities, create multiple
reflections and make the reflectivity at any given point a function of time.
Traditional techniques for reducing the RCS are appliedJ o RCS management
studies. These techniques are by covering absorber material, target shaping, and
impedance loading. The theory pertaining to absorber-covered bodies is reviewed
and the theoretical background for obscuring an image by absorber covering based on
the physical optics approximation is established. Certain types of absorbers are used
to cover metallic objects. Their effect on RCS reduction, range profile modification,
and image distortion are experimentally studied and discussed. A concept concerning
the absorber covering patterns is proposed, where the reflections from the boundaries
between the covering absorbers and conductors are also considered. By using suit­
able covering patterns the RCS can be reduced more effectively.
Some possible effects of loading an object with lumped impedances are exam­
ined. A thin straight wire is used as a test object. Plots of range profiles show that
the surface traveling wave is an important scattering mechanism of a straight wire.
Microwave images of a loaded wire are both numerically and experimentally
obtained. The specific scattering mechanism of the thin wire makes its image unique
in appearance. It is found that nonlinear loading and time varying loading can cause
the spectra of the back-scattered field to spread as compared to the spectrum of the
incident wave, which may produce several interesting phenomena, such as providing
a false doppler frequency, making the receiver unable to phase lock to the frequency
of the incident wave, generating unexpected peaks in the range profiles, and distort­
ing the reconstructed image.
-v-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table of Contents
Acknowledgements.................................................................................................
iii
A b stra c t...................................................................................................................
iv
Table of C ontents...................................................................................................
vi
List of T a b le s..........................................................................................................
ix
Ltet of Figures ........................................................................................................
x
Chapter 1 In troduction........................................................................................
1
1.1 Overview of the Dissertation .......................................................................
6
1.2 Contribution...................................................................................................
8
Chapter 2 Basic Scattering Theory of a Perfectly Conducting Object
10
2.1 Integral Equation Solutions...........................................................................
11
2.2 Approximation Solution ................................................................................
15
2.2.1 Geometrical Optics ..................................................................................
16
2.2.2 Physical Optics ........................................................................................
19
2.2.3 Geometrical Theory of Diffraction .........................................................
21
2.3 Scattered Field of Some Simple Shaped Objects........................................
24
2.3.1 Sphere.......................................................................................................
26
2.3.2 Cylinder....................................................................................................
28
2.3.3 Wedge with Finite Length......................................................................
30
2.3.4 Finite Plate ...............................................................................................
32
2.3.5 Comer Reflector......................................................................................
37
2.4 Scattering Mechanisms of a Complex Shaped O bject................................
38
2.4.1 Specular points........................................................................................
39
2.4.2 Surface Discontinuities...........................................................................
41
2.4.3 Surface Derivative Discontinuites..........................................................
41
2.4.4 Creeping Waves ......................................................................................
42
- vi -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.4.5 Surface Travelling Waves .......................................................................
43
2.4.6 Multiple Scattering .................................................................................
44
2.5 Polarization Effect ........................................................................................
44
2.6 RCS Measurement........................................................................................
48
Chapter 3 Image Understanding in Microwave Diversity Imaging .............
54
3.1 Image Reconstruction of Point Scatterers ...................................................
55
3.2 Formulation of Microwave Imaging for Conducting Objects Based on
the Physical Optics Approximation .......................................................................
61
3.3 Interpretation of the Reconstructed Images..................................................
64
3.4 Examples.......................................................................................................
71
3.5 Discussions ...................................................................................................
81
Chapter 4 Extrapolation of Available Data into Missing Bands and Its
Application to Radar Imaging ............................................................................
83
4.1
The New Iterative algorithm for Extrapolation of Data Available in
Multiple Restricted Regions ...................................................................................
86
4.2 Results...........................................................................................................
90
4.2.1 Range Profiles.........................................................................................
90
4.2.2 Reconstructed Im ages.............................................................................
99
4.3 Discussion.....................................................................................................
102
Chapter 5. Radar Cross Section Management Employing Microwave
Diversity Im aging..................................................................................................
105
5.1 Diaphanization ..................
106
5.2 Radar Cross Section Reduction ...................................................................
109
5.3 Effect of Gaussian Noise on Microwave Diversity Imaging......................
114
5.4 Fundamental Concepts for Distorting an Image .........................................
116
5.5 Discussions ........................................
123
Chapter 6. Diaphanization by Absorber Covering and Target Shaping
...................................................................................................................................
124
6.1 Absorber-covered Bodies -- Theoretical Considerations ............................
126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.1.1 Semi-Infinite Plate ...................................................................................
127
6.1.2 Physical Optics Approach for an Absorber-Covered Metallic
O bject.......................................................................................................................
129
6.2 Radar Absorbing Materials (RAM) .............................................................
136
6.3 Experimental Results for Absorber-Covered Bodies ..................................
138
6.3.1 Broadband Absorber-Covered Bodies....................................................
140
6.3.2 Surface Current Absorber-Covered Bodies ...........................................
146
6.4 Target Shaping...............................................................................................
156
6.5 Combination of Absorber Covering and Shaping.......................................
165
Chapter 7 Diaphanization by Impedance L o ad in g.........................................
173
7.1 Linear Impedance Loading ..........................................................................
175
7.1.1 Analysis of a Loaded N-port Scatterer..................................................
176
7.1.2 Formulation of a Straight Wire with Impedance Loading....................
181
........
187
7.2 Nonlinear Impedance Loading...............................................................
200
7.1.3 Scattered Fields and Images of a Thin Rod Scatterer
7.2.1 Analysis of an N-port Nonlinear Loaded Scatterer...............................
205
7.2.2 Scattering Properties of a Nonlinearly Loaded Thin Rod Scatterer
208
7.3 Time Varying Loading.................................................................................
214
7.4 Discussion......................................................................................................
224
Chapter 8 Conclusion .........................................................................................
228
8.1 Future Research ............................................................................................
230
Appendix .................................................................................................................
232
Bibliography ...........................................................................................................
235
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Tables
5.1
Effect of a Dominant Scatterer ...............................................................
113
5.2
Effect of Working Harder on the Dominant Scatterer ...........................
113
5.3
Effect Selective Elimination of Scatterer................................................
113
5.4
Effect of Reduction When All Have the Same Amplitude ...................
113
6.1
Characteristics of the GDS absorbers......................................................
149
6.2
Hierarchy of Scattering Shapes ...............................................................
158
-ix-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Figures
2.1
An astigmatic ray tube.....................................................................................
17
2.2
Reflection at a curved surface.......................
18
2.3
Geometry of a back-scattering problem..........................................................
20
2.4
Geometry for a three-dimensional diffraction problem..................................
23
2.5
Diffraction coefficient versus azimuth angle for various included
angles.........................................................................................................................
25
2.6
Diffraction coefficient versus diffraction angles.............................................
26
2.7
Normalized RCS of a conducting sphere versus k a ......................................
28
2.8
Geometry of a finite wedge diffraction problem............................................
31
2.9
Plot of back diffraction fields versus elevation angle....................................
33
2.10 Plot of back diffraction field versus azimuth angle.......................................
34
2.11 An arbitrary polygonal plate with N edges....................................................
35
2.12 Procedures of calculating the diffraction field of an arbitrary edge
37
2.13 Geometry of a dihedral comer reflector.........................................................
38
2.14 Creeping wave concept of diffraction by a curved surface...........................
43
2.15 Diffraction coefficients of an infinite edge vesus azimuth angle..................
46
2.16 Block diagram of the RCS measurement system...........................................
51
3.1. Simplified imaging geometry projected to the X - Y plane............................
56
3.2
An example of target range reflectivity together with its successive
differentials................................................................................................................
68
3.3
Positions of the scattering centers and images of a cylinder.........................
73
3.4 Geometry and reconstructed images of a hexagonal plate...........................
75
3.5
Differential range of multiple reflection path versus rotation angle and
simulated images ofa dual-cylinder object .............................................................
3.6
76
Range profiles and reconstructed images of a conducting sphere
obtained by various spectral windows...................................................................
78
3.7
80
Sketch of the B-52 airplane and the reconstructed images..........................
- x-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.1
Available data in multiple regions..................................................................
4.2
Schematic diagram of the new iterative extrapolation method pro­
posed..........................................................................................................................
87
89
4.3 Comparison of extrapolation errors and range profiles for different
various extrapolation methods with spectral window f mjn = 6 GHz to
/max = 16 GHz.........................................................................................................
4.4
92
Comparison of extrapolation errors and range profiles for different
various extrapolation methods with spectral window / mm = 6 GHz to
f max = ^0 GHz........................................................................................................
94
4.5 Comparison of extrapolation errors and range profiles for different
various extrapolation methods with spectral window f ^ = 6 GHz to
f max - 12 GHz.........................................................................................................
95
4.6 Measured fields, extrapolation errors, and range profiles of a scale
model B-52 at the broadside..................................................................................
97
4.7 Measured fields, extrapolation errors, and range profiles of a scale
model B-52 at the view line #64 ...........................................................................
98
4.8 Measured fields, extrapolation errors, and range profiles of a scale
model B-52 at the view line #85 ...........................................................................
100
4.9 Images of the metallic scale model B-52 aircraft reconstructed from
various spectral bands.............................................................................................
101
4.10 Reconstructed images of the metallic scale model of a B-52 aircraft
using cross-polarized waves.....................................................................................
102
5.1
Image of the scale model B-52 before and after diaphanization.................
Ill
5.2 Probability of detection for a sine wave in noise as a function of the
S/N and the probability of false alarm....................................................................
115
5.3
Effect of Gaussian noise on the signals, range profiles and images.
121
6.1
Geometry of an absorber-covered body illuminated by a plane wave.
126
6.2
Geometry of a semi-infinite conducting plate covered with a homo­
geneous plane layer..................................................................................................
- xi -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
128
6.3
Reflection coefficients Rp and Rv of a semi-infinite plate covered with
an absorber layer versus frequency.........................................................................
130
6.4 Magnitude and phase of Rv and Rp versus incident angle............................
131
6.5 The two equivalent structures of an absorber-covered body illuminated
by a plane wave.........................................................................................................
133
6.6 Geometries of the measurement arrangement of a conducting plate.
...................................................................................................................................
139
6.7
Typical performance of the Emerson product AN absorbers........................
140
6.8
Attenuation in dB down from metal plate when covered with absorber
AN72 in normal way and inside out........................................................................
141
6.9 Measured RCS pattern, sinograms and reconstructed images of a con­
ducting plate before and after diaphanization with geometry 1.............................
145
6.10 Mean RCS patterns, sinograms, and reconstructed images of the B-52
airplane before and after diaphanization..................................................................
148
6.11 Covering configurations of a plate and their mean RCS patterns, range
profiles and reconstructed images............................................................................
152
6.12 Geometry, mean RCS, sinogram, and reconstructed image of a tube.
...................................................................................................................................
155
6.13 Geometry and mean RCS patterns of a serrated plate using geometry
1
162
6.14 Mean RCS and images of a flat plate and a serrated plate using
geometry 2.................................................................................................................
164
6.15 Geometry, mean RCS patterns, and range profiles of triangular plates
with sharp tips and round comers arranged in geometry 2....................................
166
6.16 Picture, mean RCS patterns, and images of a scale model B-l air­
plane...........................................................................................................................
167
6.17 Covered patterns, mean RCS patterns and reconstructed image of a
plate arranged in geometry 2 with tilted angle 0 = 90°..........................................
169
6.18 Sketch of the absorber covered portions of the scale model B-l and
its RCS and reconstructed images before and after diaphanization.......................
171
7.1
176
Equivalent network of an N-port loaded scatterer..........................................
- xii -
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
7.2
Geometry of astraight wire.............................................................................
7.3
N segment loaded wire and its equivalent circuit for the m th port.
...................................................................................................................................
7.4
185
Piecewise sinusoidal expansion function and set of overlapping piece-
wise sinusoidal expansionfunctions..........................................................................
7.5
182
188
Extreme values and no-load values of the back-scattered field of a
straight wire scattered versus length in wavelength at various angles of
incidence....................................................................................................................
191
7.6 Impedance required to nullify the back-scattered fields for angle of
incidence equal to 9=45°..................................................................
192
7.7
Required passive impedance to maximize and to minimize the back-
scattered fields for angle of incidence equal to 0=45°............................................
192
7.8 Extreme values and no-load values of the broadside back-scattered
fields of a straight wire versus length in terms of wavelength.............................
193
7.9
A straight wire illuminated by an impulsive plane wave............................
195
7.10 The numerical range profiles of a straight wire at various angles of
incidence ..................................................................................................................
197
7.11 The experimental range profiles of a straight wire at various angles of
incidence ..................................................................................................................
198
7.12 Numerical and experimental fringe patterns and images of a thin rod.
...................................................................................................................................
199
7.13 Numerical range profiles of a straight wire with three loading points
at various angles of incidence.................................................................................
201
7.14 Experimental range profiles of a straight wire with three loading
points at various angles of incidence......................................................................
202
7.15 Numerical and experimental fringe patterns and images of a thin rod
with three loading points..........................................................................................
203
7.16 Comparison of the measured mean RCS patterns of the thin rod with
and without gaps........................................................................................................
204
7.17 Equivalent network of an N-port nonlinear loaded scatterer.........................
207
7.18 Time domain fields and their spectra of a nonlinearly and linearly
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
loaded wire scatterer..................................................................................................
212
7.19 Time domain scattered fields and spectra of a nonlinearly loaded wire
scatterer with three loading points............................................................................
213
7.20 Spectra of a time varying loaded scatterer.....................................................
217
7.21 The magnitude and phase of the back-scattered field and the range
profile of the loaded straight wire.............................................................................
220
7.22 The magnitude and phase of the back-scattered field, and the range
profile of the loaded straight wire, each impedance is randomly resistively
varied between (0, 100)0..........................................................................................
221
7.23 The magnitude and phase of the back-scattered field, and the range
profile of the loaded straight wire, each impedance is reactively randomly
varied between (-7 50, y'50)Q....................................................................................
222
7.24 The range profiles of a randomly loaded straight wire.................................
223
7.25 Real part and imaginary part of the polar hologram of the randomly
reactively loaded straight wire with nL = 7............................................................
224
7.26 Optical image and microwave image of a wire model airplane .................
226
7.27 A loaded conducting wedge............................................................................
226
A l.
234
Multiple reflections of a two-cylinder object.................................................
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 1
INTRODUCTION
T h e inverse or indirect scattering problem is to deduce features of source or scatter­
ing objects from the emitted or scattered radiation that has propagated to a detector
[1]. The characteristic descriptors of a scattering object are estimated from experi­
mental data, utilizing the laws that relate these characteristic parameters to the experi­
mental data in a given situation. In the electromagnetic inverse scattering problem,
the size, shape, and constitutive characteristics of an unknown scattering target are
usually the parameters to be recovered based on the knowledge of the incident field
and the resulting scattered field data [2]. Imaging is a technique to solve inverse
scattering problem. The shapes, or the constitutive characteristics of objects are
presented by pictorial representation or images.
One of the main design objectives of an imaging system is to have good image
resolution. The available recording aperture conventionally restricts the resolution
obtainable. This constraint is known as the Rayleigh resolution criterion. The basic
approach to improving resolution in microwave imaging system is to extend the
effective area of the physical recording aperture through clever data acquisition
schemes. These schemes are called synthetic aperture systems. The synthetic aper­
ture systems are usually categorized according to the relative motion between the
recording systems and the object and can be grouped into synthetic aperture radar
(SAR) [3,4], inverse synthetic aperture radar (inverse SAR), and spotlight-mode syn­
thetic aperture radar (spotlight-mode SAR) [5,6,7]. In SAR, the data are collected by
means of an airborne or space-borne radar which illuminates the target area from
different aspect angles and the position of the antenna remains fixed relative to the
flying vehicle. In inverse SAR, the recording antenna is stationary while the target is
moving to synthesize the effective aperture. The spotlight-mode SAR steers the phy­
sical antenna so that the same target area remains illuminated during the data collec­
tion interval. In the above synthetic aperture systems, the angular resolution or the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
2
-
cross-range resolution of the image is determined by the effective aperture syn­
thesized in terms of the wavelength, while the range resolution is determined by the
pulse width of the signal or the equivalent bandwidth of the signal.
The concept of synthesizing an imaging aperture by frequency sweeping was
first analyzed in [8]. It was found that a one-dimensional aperture can be synthesized
with a single coherent receiver by sweeping the frequency of the incident illumina­
tion under certain circumstances. Additionally, a two-dimensional imaging aperture
can be synthesized by a linear array of coherent receivers with one dimension of the
aperture frequency synthesized. The use of frequency sweeping or the frequency
diversity technique to achieve super-resolution in the imaging of three-dimensional
perfectly conducting objects was studied and demonstrated by computer simulation
[9]. By utilizing the intrinsic freedom of an imaging system, extremely highresolution projective and tomographic imaging of 3-D perfectly conducting objects
has been achieved by combining angular, spectral, and polarization diversities
[10,11]. In the frequency diversity imaging systems above, the object is seated on a
rotating pedestal controlled by a computer.
The resolution limit over a given
specified angular aperture and spectral band has been analyzed in [9,11,12]. The
advantage of broad-band illumination over monochromatic illumination was discussed
in [8,13].
These advantages include the avoidance of object resonances [14],
increased information about the geometrical and material properties of a scattering
object, and the reduction of coherent image artifacts. Multi-frequency imaging tech­
niques have also been applied in many other areas [15,16,17,18].
The appearance of an image depends on the scattering properties of the object
and the imaging technique utilized. The images obtained by visible light, infrared
light, microwaves, and acoustical waves all look different [19]. Image interpretation
is defined as the act of examining images for the purpose of identifying objects and
judging their significance [20]. Interpretation is an essential process through which
information can be retrieved from the images. However, interpretation of the images
should be based on the understanding of the laws that govern the input (the incident
field) and the output (the reconstructed image).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
3
-
The targets of interest in this dissertation are metallic objects. The scattering
properties of metallic objects at all frequency regimes have been extensively analyzed
[21]. The frequencies are usually divided into three regions according to the charac­
teristic dimensions of the object. These are: a low frequency region, a resonant fre­
quency region, and a high frequency region. In microwave imaging applications, the
high frequency region is most interesting. At high frequencies, the scattering prob­
lems are usually analyzed by approximation approaches, which include geometrical
optics (GO) physical optics (PO), and the geometrical theory of diffraction (GTD)
[21]. To relate the geometrical shape of a conducting body to its scattered field, the
PO approximation is usually adopted. It was shown by Bojarski [22] and Lewis [23
], that a three-dimensional Fourier transform relationship exist.: between the shape of
a perfectly conducting object and its back-scattered far field under the PO approxima­
tion. The shape of a convex body can be retrieved if the back-scattered field can be
measured for all frequencies and all directions. However, the spectral and angular
windows for the data are usually restricted by practical constraints. Radar people
model the object with a target reflectivity function [24], which is defined in terms of
a range coordinate and a cross-range coordinate measured from the center of the syn­
thetic aperture. Basically, the target reflectivity function is also derived from the PO
approximation. The principle of frequency diversity imaging systems [9] is also
analyzed based on the PO approximation. An object scattering function was first
defined and then it was shown that the object scattering function and the range
corrected field are a 3-D Fourier transform pair [10,11].
However, the PO approximation is inadequate for scattering problems where
edge diffractions, multiple reflections, creeping waves, and traveling waves are the
dominant scattering mechanisms. The above scattering mechanisms are important
contributors to the scattered field of a complex shaped conducting object [21].
The microwave image of an object also depends on the imaging technique util­
ized. Different data acquisition schemes may produce different images. The imaging
technique used in this dissertation is the microwave diversity imaging scheme
achieved by combining angular, spectral and polarization diversities, developed at the
Electro-Optics and Microwave-Optics Laboratory of the University of Pennsylvania.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
4
-
There is a common property between the data acquisition arrangements of microwave
diversity imaging and the well-known X ray imaging technique, computer-aided
tomography (CAT), which has been used to obtain the image of a slice of the object
and has wide applications in medical imaging [25,26]. Both systems involve rotation
of objects. This similarity suggests that the reconstruction algorithms used in CAT
may be applied to microwave diversity imaging. The two reconstruction methods
used in CAT are the Fourier inversion method [27,28] and filtered back-projection
method [26]. These two methods can yield equivalent results [28]. Microwave
tomographic and projective images reconstructed by these two methods have been
obtained [10,11,29]. The imaging system involving rotation of the object has been
analyzed independently using the range-Doppler principle
by radar people
[6,30,31,32]. But the same scheme can be interpreted as a tomographic reconstruction
problem and can be analyzed using the projection-slice theorem from CAT [7].
Imaging workers usually model the object consisting of discrete points and
assume that each point radiates isotropically regardless of the aspect angle. Accord­
ingly, they analyze the imaging system by the point-spread function [31]. However,
a conducting object cannot be modeled in this manner. Even though the conducting
object can be viewed as a combination of discrete scattering centers at a specific
aspect angle, the locations of the scattering centers may migrate and their scattering
strength can change as the aspect angle is changed. Therefore, no fixed point-spread
function can be precisely defined in the conducting object imaging systems because
the point-spread function is space-variant and target dependent.
In this dissertation, we will investigate microwave imaging of metallic objects
employing frequency, angular, and polarization diversities, with particular emphasis
on image interpretation and prediction based on analysis of the scattering mechanisms
and the filtered back-projection algorithm utilized in image retrieval. We will ela­
borate on the connection between the various scattering mechanisms and the recon­
structed images; interpret what the images represent; and then predict what the image
will look like over given spectral and angular windows.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
5
-
After the microwave images are satisfactorily interpreted, one may retrieve valu­
able information from the images. It will be shown later that the brightness of each
image cell reflects the average scattering strength of the object over the specified
spectral and angular windows. This suggests that high resolution microwave diver­
sity imaging can be applied to the study of radar cross section (RCS) reduction.
RCS is a measure of the equivalent size of a target as seen by the radar. RCS
studies create interest because of the unique applications of RCS. A special issue on
RCS was published in the Proceedings of IEEE in 1965. Since then, the technology
has steadily advanced. However, due to the interests of national security, some
aspects of radar cross section measurement, prediction and reduction are not pub­
lished [33].
RCS is a function of aspect angle, frequency, and polarization. In some applica­
tions, it is desirable to enhance the RCS over some specified angles. On the other
hand, in other applications it is often desirable to reduce or minimize the RCS over
some spectral and angular windows to avoid target detection [21],
RCS management is a general term for obtaining the RCS of a scattering object
by manipulating the distribution and strengths of hot spots or flare spots of a target
over prescribed spectral and angular windows and states of polarization. Hot spots
represent all those portions that have major contribution to the received scattered
fields.
To reduce the RCS, the hot spots must be located and identified. Several
approaches can be used to achieve this purpose. Examples include simply engineering
guesses, numerical calculations, and experimental measurement [33]. The advent of
high resolution imaging radar can aid in the identification of scattering centers. In
this dissertation we will employ the high resolution microwave diversity imaging sys­
tems to locate and identify the hot spots of a complex object.
After the hot spots have been located, the next problem is to remove or suppress
them. Techniques of reducing RCS have been extensively analyzed in [21] and are
grouped into three classes: covering the object with absorbing material or substituting
the metallic material with resistive or composite material, shaping the target, and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
6
-
impedance loading. These techniques will be applied in this dissertation to reduce the
RCS while microwave diversity imaging will be employed as a new tool to examine
the effectiveness of the RCS reduction technique.
Radar designers try to make radars capable of detecting very weak signals and
recognizing or identifying the target. In contrast, target designers try to reduce the
target’s RCS to elude radar detection, and to disguise the target’s appearance to
impede radar recognition. From the electronic counter measures point of view, the
optimistic techniques should possess the capability of reducing a target’s RCS as well
as obscuring the target’s images. In this dissertation we will consider RCS manage­
ment from both detection and imaging view points. We will examine how the
images are changed after the traditional RCS reduction techniques are applied. We
will also propose some fundamental rules for obscuring an image based on the under­
standing of the scattering mechanisms and image formation.
Microwave diversity imaging has been extensively studied in the Electro-Optics
and Microwave-Optics Laboratory of the University of Pennsylvania employing a
unique experimental microwave imaging facility. This facility has proved to be a
very valuable tool in the understanding of electromagnetic (EM) scattering and
diffraction and the understanding of pertinent mechanisms in the formation of a
microwave image.
1.1 Overview of the Dissertation
The whole dissertation is structured into seven chapters. Each chapter is sum­
marized as follows:
The complex field of a coherently illuminated object is the basic measured
quantity for all scattering problems, because all the parameters of interest can be
derived from the scattered field. In chapter 2 the basic scattering theory of a perfect
conducting object will be reviewed, starting from the exact integral equation solution
of Maxwell’s equation followed by its approximate solution. The scattered field of
some simple objects and the scattering mechanism of a complex object will also be
summarized. The polarization properties of the scattered field contain the feature
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
7
-
information of the object. At high frequencies the scattered field can be expressed as
from a combination of some discrete components and the polarization status of each
component is independent or slowly varies with frequency. With a high resolution
radar, these discrete components can be resolved and can be considered separately.
The technique of RCS measurement and the calibration procedures will be discussed
at the end of this chapter.
In chapter 3 the image formation for microwave diversity imaging will be
explained by investigating the scattering mechanism of the object and the procedures
of the reconstruction algorithm. The images reconstructed from a specified angular
window and spectral window will be predicted and interpreted. Several typical
numerical and experimental examples will be demonstrated to show the validity of
microwave images that can be generated under various conditions. These examples
serve to illustrate the image’s dependence on data acquisition parameters and the
scattering mechanism and prove that care must be taken to achieve a proper interpre­
tation of the image.
To locate the hot spots of an object, fine resolution of the object detail in the
reconstructed image is essential. The range resolution of a frequency diversity imag­
ing system is inversely proportional to the bandwidth coverage of the measurement
system. In practical situations, due to the limitation of the measurement system or
restriction of bandwidth allocation, the observed data can lie in a narrow band or in
multiple restricted spectral regions and the bandwidth of each region can be too nar­
row to give acceptable resolution. In chapter 4 an algorithm will be used to extrapo­
late data in the missing bands from the data available and apply it to the radar imag­
ing.
In chapter 5 a new term "diaphanization" will be introduced, which is defined as
the techniques of reducing RCS and techniques of obscuring the image. The pro­
cedure of employing the microwave diversity imaging system to diaphanize a target
will be described. RCS management is treated not only from the detection point of
view but also from the image perspective. Some fundamental rules of obscuring an
image will be proposed. The effect of Gaussian noise on the reconstructed image will
also be examined.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
8
-
Diaphanization by an absorbing material covering and target shaping will be stu­
died in chapter 6. Theoretical analysis of absorber-covered bodies will be reviewed
and formulated under the physical optics approximation. Properties of radar absorb­
ing material (RAM) will be briefly described. Certain types of RAM will be used to
cover the object. Their effects on the RCS and images will be examined. A hierar­
chy of scattering shapes provides a fundamental rule for shaping a target. The RCS
of some object shapes and their frequency dependence will be experimentally stu­
died and discussed.
In chapter 7 diaphanization by impedance loading will be numerically examined
and studied. Analysis of a linearly and nonlinearly loaded N-port scatterer will be
reviewed. The moment method will be used to numerically calculate the scattered
field of a loaded straight wire scatterer and some interesting phenomena of a loaded
wire will be discussed. These phenomena include the scattering properties, the
spread spectrum of a nonlinear loading and a time varying loading, the range profiles
and the reconstructed images.
1.2 Contribution
The significant contributions of this dissertation are summarized as follows:
1.
The image from a microwave diversity imaging system is understood by investi­
gating the scattering mechanisms of the object and the image reconstruction
algorithm. The effect of the different scattering mechanism on the reconstructed
image has been established. To our knowledge such an approach has not been
reported before. Several numerical and experimental results can be satisfactorily
explained. The image of a metallic object obtained by prescribed spectral and
angular windows can be predicted accordingly. This successful interpretation is
fundamental to research in RCS management studies and target recognition.
2.
A new iterative algorithm has been devised to extrapolate the data available into
missing bands. As a consequence of extrapolation, resolution of the microwave
diversity imaging for a given bandwidth can be improved, and the system’s abil­
ity to pin-point the distribution of hot spots is increased.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
3.
9
-
A new term "diaphanization", which is defined as the techniques of reducing
RCS and techniques of obscuring the image, has been introduced. RCS manage­
ment is treated not only from the detection point of view but also from the
imaging perspective. This opens a new direction for future research in RCS
management studies.
4.
The robustness of the microwave diversity imaging to the Gaussian noises has
been demonstrated. Some fundamental rules of obscuring an image have been
proposed and experimentally studied.
5.
A way of covering metallic objects with absorbing materials, which is also con­
sidered from the absorber shaping perspective, is proposed and experimentally
studied. This technique is a combination of absorber covering and shaping.
This technique creates artificial discontinuities in surface impedances and causes
appearance of unexpected detail in the image, which might deceive the radar’s
"eye" and make it unable to recognize the object.
6.
Some phenomena of a lumped linearly and nonlinearly loaded straight wire have
been numerically and experimentally examined. These phenomena include the
scattering properties, the effectiveness of providing false Doppler frequency by
time varying loading, and the effect of fixed loading and random loading on the
range profiles and the reconstructed images. Plots of range profiles show that
the surface traveling wave is also an important scattering mechanism of a
straight wire even though the incident angle is not close to the end-on direction.
The effect of impedance loading on the reconstructed image has not been
reported before.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 10-
CHAPTER2
BASIC SCATTERING THEORY OF
A PERFECTLY CONDUCTING OBJECT
W h en an electromagnetic (EM) wave impinges on a conducting body, it induces
oscillating charges and currents on its surface. These induced charges and currents
are the sources of the scattered field. The total field at an observation point in the
presence of a conducting body is comprised of the scattered field and the incident
field, which is the field without the object and is assumed known. Therefore, the scat­
tered field is defined as the difference between the total field and the incident field.
The theoretical formulations for scattering problems rest on Maxwell’s equa­
tions. Wave equations can be derived from Maxwell’s equations. Closed form solu­
tions for wave equations are available only when the scattering geometry coincides
with one of the few separable coordinate systems. For any shape of scattering
geometries, exact integral equations can be formulated. However, these formulations
were not considered useful until the advent of computer age. Numerical techniques
can be applied to a body of arbitrary shape and generally are only limited by the
object to wavelength ratio. When the object size is large compared to a wavelength,
approximation solutions are usually used to overcome the computing difficulty.
In this chapter, we will briefly review the basic scattering theory of a perfectly
conducting object starting with Maxwell’s equations. The material to be described
will be confined to those topics which will be used later in this dissertation.
In Sec. 2.1, the integral equation solution will be formulated assuming that the
time dependence of the field is in sinusoidal harmonics. In practical situations, the
high frequency region is of great importance. Asymptotic approximation solutions
including those for geometrical optics, physical optics, and the geometrical theory of
diffraction will be summarized in Sec. 2.2. A complex shaped object can usually be
divided into several objects of simple shapes. The scattered field of some simple
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-11 -
objects will therefore be discussed in Sec. 2.3, where these simple shapes will be
used as examples in later chapters. In the high frequency region, the scattered field of
a complex shaped object can be attributed to a combination of several scattering
mechanisms. The scattering properties of each mechanism is fundamental to the
understanding of the scattering of a complex shaped body and will be discussed in
Sec. 2.4. Polarization of the scattered field carries important information about the
target. In Sec. 2.5 we will express the scattered field as a combination of some
discrete components and show that the polarization status of each component is
independent or slowly varying with frequency. Precise measurement of the scattered
field is very important in obtaining a high quality microwave image. Magnitude and
phase are both required in the measurement. In Sec. 2.6, techniques of RCS measure­
ment and the calibration procedures needed to achieve high accuracy will be dis­
cussed.
2.1 Integral Equation Solutions
Assume that all of the sources are in what is otherwise free space, Maxwell’s
equations for harmonic time dependence e
can then be expressed as,
V xfiV ) = - j ^ o f t F )
= ycoeofV) + TV)
(2.1)
(2.2)
V-£V) = p(7rf)/e0
(2.3)
V F tff) = 0
(2.4)
and the continuity equation is
?Q*) = -j(op<f)
(2.5)
where |i.0 and £q are the permeability and permittivity of free space respectively, r* is
theposition vector,
T
is the current density and p is thechargedensity. It is con-
venientto introduce the scalar potential <(>and vector potential A as
Ft(r) = — VxAV)
M-o
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.6)
- 12and
£(r>) = -V<$><r)-j
(2.7)
If the two potentials are related by the Lorentz gauge condition
V -^ V ) + j coHoEo^C^ = 0.
(2.8)
the wave equations can be written as
V t f + k 2t = - p /
(2.9)
V2<|>+ k 2((>= -p/eo
(2.10)
and
where k 2 = co2|ioeo. and k is called the wave number or propagation constant in free
space.
The solution of X can be expressed as
X = \io jf(f)G (?,?*) dv'
(2.11)
where dv' is the volume element in the source region, G
is called the Green’s
function and is the solution of the following function
V 2G + k 2G = -5(r» - r )
(2.12)
where 8 is the Dirac delta function. G is given by
G ( r r ') = T4nR
T
(2<13)
with
r
= i/?i = \ r - r \
(2.i4)
By substituting Eq.(2.13) and Eq.(2.11) into Eq.(2.7), and after laborious vector
operation, one can show that the field scattered by a conductor with surface current
density Ts (/*) is [34]
{ - y c o p /,( n + —
S’
[Ts ( r y i R]tR [ - k 2 + ^ U k + - h ]
J tutQ
j(jk + j)? sr))G (? ,
K
K
r) ds'
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.15)
- 13 and
/fa ) =
HoJ (ik + \) G
S’
K
(?, r ) f ( f ) x lR ds'
(2.16)
where,
I _
R \t-r\
=£
n 17s
(
r
}
ds' is the area element, and S' is the conductor surface.
From the above equations, it is apparent that the scattered fields at any position
can be determined if the current source distributions are exactly known. When a
metallic object is illuminated by a plane wave, current will be induced on the surface
and the induced current will become the source of the reradiated or scattered field.
However, the induced current distribution usually cannot be expressed in closed form
except for those objects whose surfaces coincide with the surfaces of orthogonal cur­
vilinear coordinates. To find the source distribution, boundary conditions have to be
applied. For the scattering problem involving a perfect conductor, the boundary con­
ditions on the conducting surface requires that
hxM101 = f i x ^ + i ^ ) = 0
(2.18)
and
n x ft101 = nx(Fts + tf inc) -
Ts
(2.19)
where l? ot and Fttot are the total electric field and the total magnetic field in the
presence of the conductor; j?s and Ffs are the scattered electric field and scattered
magnetic field due to the induced surface current, ^ inc and Jrtinc are the incident
electric field and incident magnetic field, and h is the unit outward normal to the sur­
face. As a result, we have the following integral equations
n x ^ C r ) = -2n x
j { -y c o p ^ r) +
v
<7ytR]lR [-k 1 + \ ( j k + -U ]
I u)£n
K
-T-C i k + ± ) J s r ) } G ( ? , n ds'
R ''" R
and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
K
(2.20)
- 14Ts (f) = 2n x f tinc + 2 n x | ( j k + j ) G ( f , r ) Ts ( f ) x lr ds'
(2.21)
Eq.(2.20) is called the Electric Field Integral Equation (EFEE), and Eq.(2.21) is called
the Magnetic Field Integral Equation (MFIE). Note that in Eq. (2.20) and Eq.(2.21),
a factor 2 has been multiplied to account for the presence of the perfectly conduct­
ing surface.
For an object of general shape, the solution must be based on numerical tech­
niques. Numerical techniques can be applied to a body of arbitrary shape and gen­
erally are only limited by the wavelength size of the body. The moment method is
the most popular numerical method for treating the scattering problem of a perfectly
conducting object. This method has been extensively studied [35,36,37] and will be
used to solve the scattering property of a straight wire loaded with impedance in
chap.7.
Once the surface current is determined, the scattered fields can be determined by
Eq.(2.20) and Eq.(2.2l). If only the far fields are of interest, these equations can
further be simplified. Under the far field condition, the distance R in the phase term
and denominator term can be approximated by
R ~ r - lr r*
J_ _ J_
R ~ r
where r
is the distance between the observation point and the originorthereference
point, lr
is the unit vector of the line connecting the origin and theobservation point
and is parallel to lR. The far zone vector potential X can be expressed as
A V ) = ~ r ~— f £ ( r * V Wr‘^ ds'
4nr Js'
(2.22)
Furthermore, all the terms in the integrand of Eq.(2.15) with order higher than or
equal to MR can be neglected. The far fields are then reduced to
g(i*)= ri y
*
47tr
.j
( T .c n - f i o n - h t , ] eill' r
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
<2.23)
- 15 (2.24)
It can be shown that
X is related to X by
X = - ja X j
where
XT is
the transverse component of
(2.25)
X
along the direction lr . Eq.(2.23) and
Eq.(2.24) will be the fundamental equations used to calculate the far field of a con­
ducting scatterer.
The previous analyses are based on the time harmonics assumption. Current
developments in high resolution radar and EMP technology have created interest in
the radiation and scattering of transient waveforms from conducting bodies of various
types. Transient scattering gives more insight into the scattering mechanism of a
scatterer [38]. There are two independent techniques for solving transient scattering
problems. The first involves the computation of the frequency domain response of the
structure, which is then Fourier transformed to yield the time-domain response. The
second is based on a direct formulation of the integral equation in the time domain
and its subsequent solution using a time-stepping rather than matrix inversion algo­
rithm [39]. In chap.7 the transient behavior and the spread spectrum behavior of a
straight wire scatterer with nonlinear loading will be studied. The first of the two
techniques mentioned above will be used. Computational aspects of these two
methods have been summarized in [37,40].
2.2 Approximation Solutions
As the dimension of an object gets larger, the solution obtained by a series
expansion becomes poorly convergent. Additionally, the moment method is restricted
to objects with dimensions not large in terms of a wavelength. The high frequency
approach or the asymptotic approximation method can then be used to overcome
computer limitation.
When an object is large compared to a wavelength, the scattering and diffraction
are found to be essentially a local property of the object [41]. Examples of local
scattering regions include points of specular reflection, shadow boundaries, tips,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 16 -
edges, etc. The high frequency approximations to be discussed are geometrical optics,
physical optics, and the geometrical theory of diffraction.
2.2.1 Geometrical Optics
Geometrical optics was originally developed to analyze the propagation of light
where the frequency is so high that the wave nature of light need not be considered.
In this approximation, the optical laws may be formulated in the language of
geometry'. The energy may then be regarded as being transported along certain curves
or light rays [42].
Some of the principal properties of a geometrical field are [43]:
1.
The geometrical rays may be defined as a family of curves which are every­
where normal to the constant geometrical wavefronts. In a homogeneous
medium, all rays are straight lines. If the refractive index is not a constant, the
rays are curved, and the rays always bend toward the regions of higher refrac­
tive index.
2.
The variation of the amplitude of the geometrical optics field within a ray tube
is determined by the law of energy conservation. For example, an astigmatic
ray tube is shown in Fig. 2.1. The principle radii of curvature of d a 0 are
p! and p2, while the principle radii of curvature of d a are (pj+S) and (p2+S).
The law of energy conservation requires that
da0
da
P1P2
(Pi+S)(p 2+S)
(2.26)
The magnitude of the electric field £ at d a can then be expressed in terms of
the field £ 0 at d a 0 by the relation
(2.27)
(Pi+S)(P 2+S)
3.
The phase delay along a ray is given by the optical length of the ray. The phase
delay along all rays connecting any two equi-phase surfaces is the same.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 17 -
line caustic
2
Fig. 2.1 An astigmatic ray tube. [44]
Combining 2 and 3, the geometrical optics field Z? at S can be written as
*<S) = m
e - ^ y j
(pij ; ; ^ 5) . * •
(2.28)
where \j/(0) is the phase at d Cq.
It is important to find an expression for the radii of curvature of the wavefront
of the reflected wave in terms of the geometrical radii of curvature of the object sur­
face and the parameters of the incident waves. Reflection at a curved surface is
shown in Fig. 2.2. The geometrical optics reflected field at a distance 5 away from
the reflection point can be expressed as [41]
t ( S ) = El 1(Qr )R A / - ■- Pl' P2 —
V (pf+S)(p 2'+S)
(2. 29)
in which p f and p£ are the principal radii of curvature of the reflected wavefront at
the reflection point QR, £f* (QR) is the incident field at QR, and R is a dyadic with
R =e'p e rp - el e[
(2.30)
where e rv , el are unit vectors perpendicular to the plane of incidence, and e lp , e rp are
unit vectors parallel to the plane of incidence so that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 18 -
ep = ev xs
(2.31)
where £ is the unit vector in the direction of propagation.
*k
Fig. 2.2 Reflection at a curved surface[41].
The general relationship between pf, p£ and the parameters of the surface and
the incident wave can be found in [41]. For the two dimensional case, i.e., an infinite
line source parallel to the axis of a convex cylinder of arbitrary cross-section, pr can
be expressed as [44],
— = — + ----pr
p‘
r CCOS0;
(2.32)
where p' is the principal radius of curvature of the incident wavefront at QR, and r c
is the radius of curvature of the surface at QR.
The geometrical optics field may require no correction when to is sufficiently
large. This is the case for back-scattering from smooth curved surfaces with radii of
curvature very large in terms of a wavelength. However, geometrical optics fails in
the case of a cylinder and flat plate, where one of or both of the principal radii of the
surface approach infinity. It also fails when the specular point approaches a shadow
boundary on the surface.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 19-
2.2.2 Physical Optics
Physical optics is an approximation that expresses the induced surface current
density in terms of the incident field. When the object dimensions are much larger
than a wavelength, physical optics assumes that the field at the surface of the scatter­
ing body is the geometrical surface field. This implies that, at each point on the
illuminated side of the scatterer, the scattering takes place as if there were a fictitious
infinite tangent plane at that point. While over the shadow region, the field at the sur­
face is zero. For a perfect conductor, the surface current is assumed to be
*
n x f t tot (ry) = 2/1 ( f') x ff1(7*) 7*on the illuminated surface
Z<r> = ' 0
elsewhere
(2.33)
Let the incident plane wave at the differential surface ds' be
& < ?) = it° e jldi r
where
it0 is
(2.33A)
the incident field at the origin and is a constant vectorperpendicular to
thedirection of incidence, and lt is the unit vector in the
direction of incidence.
Substituting the above expressions into Eq.(2.23) and (2.24), one has
A*(r>) = — \fi ( O x t f V j e jk ^
2 n rJ
& (? >
=
2nr
-
yr ds
J { /iC ^ x tfV ) ~
J
Jill
(2.34A)
[ n ^ x F P r y i r V , )
■eik{f '~ ^ r ds'
ff* (j*) - J h L ^ L f ^ (j*) x
2 nr D
sai
(2.34)
x lrejk{lr~ii)r ds'
(2.35)
where s m refers to the illuminated surface.
If the surface can be expressed as z' = / {x ,y'), the normal vector to the surface,
n , and the element area, d s', on the surface become
a - - 0 / f ix ' )* - O f f iy ' )y+ ?
V i + o / / a / ) 2+ o / / a / ) 2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.36)
-20and
ds' = dx'dy' V1 + 0 / fix' )2+(df Idy' )2
(2.37)
The general expressions of the far electric and magnetic fields can be obtained by
substituting Eqs. (2.36) and (2.37) into Eqs. (2.34) and (2.35).
Consider the special case of back-scattering, where /) = - l r . Without loss in
generality, assume that the z axis is along the lr direction, and the geometry is as
shown in Fig. 2.3. It can be proved then that Eq. (2.34) and Eq.(2.35) become
Eis( f) = Jkf jkr ^ f (f -a (r^)exp( - j 2kz' )ds'
2tcr
aJm
= ~1k2^
(z n OOJexp( - j 2kz' )ds'
antenna
T
z= L
\z -z
- shadow
boundary
shadowed side
Fig. 2.3 Geometry of a back-scattering problem.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.38)
(2.39)
-21 -
It is noted that £* and Z?°, Fts and
Ft0 are in the
same directions. Furthermore,
Cn-z)ds'=nzds' is the projection of the element of the surface area ds' to the x '- y '
plane; hence, nz ds' =ds'z =(ds'z Idz' )dz' where s'z is defined as the projection of the
surface area for z>z' onto the x' —y' plane. Therefore, Eq.(2.38) and Eq.(2.39)
become
& <T) =
| c * v H 2 k z ') ^ d z '
(2.40)
and
Ft*r)=
J exp(-y 2kz' )~ ^~ d z'
(2.41)
where L is the maximum length of the object in the z direction, and the xf - y'
plane is taken to be behind the shadow boundary (we have defined s'z=0 for z' > L
and s'z = s 'max for z' < 0). Note that the scattered field is proportional in magnitude
to the Fourier transform of ds'z Idz' .
It is worth noting also that the equations obtained from physical optics for the
scattered field from a conducting body often reduce to the equation of geometrical
optics in the high frequency limit.
2.2.3 Geometrical Theory of Diffraction
Neither Physical optics nor geometrical optics can predict a nonzero field in the
shadow region. However, geometrical optics may be extended to include a class of
rays, called diffracted rays, which permits the calculation of fields in the shadow
region of a scatterer. This theory is called geometrical theory of diffraction (GTD)
and was developed by Keller [45].
In 1953, Keller introduced diffraction coefficients and related them to the
asymptotic expression developed by Sommerfeld. Keller made the following postula­
tions [45]:
1.
The diffracted field propagates along ray paths that include points on the boun­
dary surfaces. These ray paths obey the principle of Fermat, known also as the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 22 -
principle of shortest optical path.
2.
Diffraction, like reflection and transmission, is a local phenomenon at high fre­
quency. That is, it depends only on the nature of the boundary surface and the
incident field in the immediate neighborhood of the point of diffraction.
3.
A diffracted wave propagates along its ray path so that
(a). Power is conserved in a tube of rays, and
(b). Phase delay equals the wave number times the distance along the ray path.
The diffracted field can be written by a symbolic expression
[ E d ] = [ D 1 [ E l ] A Cs)e~Jks
(2.42)
where [ E d ] and [ E l ] are column matrices consisting of the scalar components of
the diffracted and incident field respectively, [ D ] is a square matrix of the scalar
diffraction coefficients, s is the distance from the observation point to the diffraction
point, and A (s) is the spreading factor.
Shown in Fig. 2.4 is the geometry for a three-dimensional diffraction problem,
where s' is the source point, s is the observation point, (2-n )k ( 0 <, n <2 ) is the
included angle of the wedge, and Q is the diffraction point which is determined by
the Fermat principle. It is more convenient to describe the diffraction system using
ray-fixed spherical coordinates with Q as the center. The plane containing the
incident ray and the edge of the wedge will be referred to as the plane of incidence;
while the plane containing the diffracted ray and the edge will be referred to as the
plane of diffraction. The unit vector s ' is in the direction of incidence and the unit
vector s is in the direction of diffraction. Using this ray-fixed coordinate system,
E}{s)
/•—
s1
* .
0
o
i
1
Eq.(2.42) can be expressed as
E ‘'(Q)
-D v
EH Q )
A (s)e ^
(2.43)
where p ' and v' refer to directions parallel and perpendicular to the plane of
incidence respectively, and p and v refer to directions parallel and perpendicular to
the plane of diffraction respectively. Keller’s diffraction coefficient may then be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-23-
written as
plane l o f
diffra< t i o n
plane o f in c id e n c e
Fig. 2.4 Geometry for a three-dimensional diffraction problem.
^
P
•
e jW V sin(7t/rt)
B)2 rts i„ 9 '
1
cos 7t/n - cos (<j>-<t)')/n
1
cosrc/n - cos(<jH-<J>,)/rt
(2.44)
The above expression is valid when the observation point is apart from the shadow or
reflection boundaries, and will become singular as a shadow or reflection boundary is
approached.
To overcome the singular property of the Keller’s diffraction coefficient,
Kouyoumjian and Pathak [41] proposed a uniform theory of wedge diffraction,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-24-
known as UTD. The UTD can be used to treat different types of waves, including
plane waves, cylindrical waves, conical waves and spherical waves, while avoiding
the singularity problem near the boundary regions. However, the UTD formulas are
much more complicated. The formula in the UTD requires calculating the diffraction
coefficient and the spreading factor can be found in [41]. The computer program used
for calculating the diffracted field using the UTD has been given in the appendix of
Ref [44].
Edge diffractions will become the major contributor to the scattered field of a
complex shaped object in the absence of specular points. It will be very helpful if the
diffraction coefficients can be numerically plotted. In the following, all the diffraction
coefficients were calculated using the UTD. For an infinite wedge, back-scattered
fields exist only when 6 = 90°. The plots of Dp and Dv versus the azimuth angle
<|> (<M>0 for various exterior angles n are shown in Fig. 2.5. From these figures, one
can observe the strong dependence of the diffraction coefficient on the azimuth angle.
This is the main difference between the diffracted fields due to an edge and the scat­
tered fields of a thin wire. The back-scattered field of a thin wire is independent of
the azimuth angle. It is also noted that the dependence of the diffraction coefficients
on the exterior angle is very small except at those angles approach to <|>= n n.
If the incident angle is fixed and the diffraction angle is varied, i.e., the bistatic
angle is changed, the diffraction coefficients Dp , Dv versus the diffraction angle for a
given set of parameter (n=2, 0, =0d=9O°) are shown in Fig. 2.6.
2.3 Scattered Field of Some Simple Shaped Objects
A complex shaped object can usually be divided into several parts of simple
shapes. It is important to know the RCS properties of some commonly encountered
simple objects. In the following, the scattered field of some simple objects will be
summarized.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-25-
diffrac. coef. in dB
20
10
•
-10
■
-20
45
90
angle in degree
(a)
45
90
135
180
135
180
135
180
diffrac. coef. in dB
0
-20
0
diffrac. coef. in dB
(b)
-10
-20
45
90
angle in degree
(c)
0
Fig. 2.5
Diffraction coefficient versus azimuth angle,
(a), n =2.0, (b). n=1.6, (c). n=1.4.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0
45
90
135
180
angle in degree
Fig. 2.6 Diffraction coefficient versus diffraction angles, where
0i = Qd = 90°, n=2.
2.3.1 Sphere
Consider a conducting sphere with radius a being illuminated by a plane wave.
The wave is incident in the - z direction and the incident field If* is polarized in the
x direction. In this geometry, the homogeneous Helmholtz equation of Eq. (2.9) can
be written in spherical coordinate form, and its eigenfunctions are called the spherical
wave functions. A plane wave can be expressed in terms of the spherical wave func­
tions [46]. The scattered field can then be expressed as a superposition of the infinite
spherical wave functions. Coefficients of the infinite series expressions are deter­
mined by the boundary conditions which are E Q(a ,9,<j)) = £^(0 ,0,0) = 0.
The scattered field can be expressed in terms of 0 and <]> components. For the
special case of back-scattering, (i.e., 0=0°,0=0°), the far back-scattered field can be
expressed as
£* = E q
s 0 = Exx
(2.45)
2n+ l
with
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-27a
n
B
n
where
= f / y» 2n+l j *(ka)
K J) » 0| + 1) h ? \k a )
(2.46)
( / r *1 2» +1 W n W Y
n (n + l) [lcah£l\ka )]'
differentiation,
denoted
with
a
(2.47)
prime,
is
with
respect
ka ,
to
h^l\ x ) = j n(x) + jyn(x), where j n(x) and yn(x) are the spherical Bessel functions of
the first and second kinds with order n, respectively. The radar cross section is,
(2.48)
To numerically calculate the RCS of a sphere, it is not possible and not practical
to calculate the infinite series. The required number of terms to calculate the RCS of
a sphere has been discussed in [21]. However, if
is very large, the convergence
rate will be very slow, and the numerical calculation of the Hankel function and its
derivative becomes difficult because more significant digits are required. In that case,
an asymptotic expression will be more convenient [21].
The RCS of a sphere normalized with respect to its geometrical cross section
versus ka is shown in Fig. 2.7. The frequency regions are usually divided into three
areas according to the value of ka. According to [21], the low frequency region (or
Rayleigh region) involves ka < 0.4; the high frequency region (or geometric optics
region ) involves ka > 20; and the resonant region involves 0.4 < ka < 20. In low
and high frequency regions, an approximate solution involving a very simple expres­
sion [21] can replace the series expansion solution. The normalized RCS is propor­
tional to (ka)4 for ka<0A and approaches a constant value for ka>20. In the resonant
region, the oscillatory nature is due to the interference of the direct reflection wave
and the "creeping wave" that circles the rear of the sphere and is launched back in
the direction of the radar. The contribution of the creeping wave decreases as the fre­
quency increases, which can be seen by the damped interference pattern. The effect
of creeping waves on the reconstructed image will be discussed in a later chapter.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-28 -
4
(N
*
3
CO
u
a*
2
•8
N
1
o
s
5
0
10
k tfi
Fig. 2.7 Normalized RCS of a conducting sphere versus ka .
2.3.2 Cylinder
For the geometry of an infinite cylinder, Eq.(2.9) can be written in cylindrical
coordinates and the eigenfunctions are called cylindrical wave functions. A plane
wave can also be expressed in terms of cylindrical wave functions [46]. Consider an
infinitely long circular cylinder with an axis in the z direction being illuminated by a
plane wave. The arbitrarily polarized incident wave can be resolved into two com­
ponents: a TE component with the E-field parallel to the cylinder axis and a TM
component with the H-field parallel to the same axis. In this two-dimensional case,
the axial incident electric field will produce only an axial scattered E field, while the
axial incident H field will produce an axial H field. Therefore, the scattered field of
an arbitrary polarized incident wave can be found by solving the scattered field from
the
TE
and
TM
components
separately.
The
boundary
condition
is
Ez (p=a) = E^(p=a) = 0. The scattered far field for the two polarization cases are
'gjikp-n/4)
oo
^
Jn(ka)
E STE — £ l TE
v 7C
V£p
Bt l
H ? \k a )
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.49)
-29i(ikp-Ji/4)
cos/t<)>
(2.50)
where primes indicate the derivative with respect to k a , and
1
e«= 2
n=0
n *0
(2.51)
The Hankel function is a linear combination of the Bessel functions of the first kind
and second kinds,
H « \k a ) = Jnm
+ iYn(ka)
(2.52)
In the above equations, p is the distance from the cylinder axis to the observation
point, and <j>is the bistatic angle.
The scattering property of an infinite cylinder is usually expressed as scattering
width, which is defined as the RCS per unit length. The scattering width of the two
cases are
(2.53)
COSH 0 I
COSn
I
(2.54)
The number of terms required to numerically calculate the RCS of an infinite
cylinder has been discussed in [21]. For ka c l and ka"> 1, Eq.<2.49) and (2.50) can
be substituted by an approximate expression to avoid the difficulty of numerical
evaluation.
In practice, the objects of interest are of finite dimension. For the finite case,
exact analytic expressions are not available, and one should resort to approximate
solutions. If the dimension of the object is large compared to a wavelength, the scat­
tered fields are usually found by physical optics approximation. The expressions can
be found in [47].
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-30-
2.3.3 Wedge with Finite Length
For the case of an infinite conducting wedge, the diffracted field exists only in
the forward direction, i.e, 0 = tc- 0 ' (see Fig. 2.8). However, if the length of the
wedge is finite, the diffracted field can be observed in any arbitrary direction due to
the effect of finite length.
The diffracted field of a finite edge for arbitrary incident and diffracted angles
has been treated [48], where the concept of equivalent electrical current and
equivalent magnetic current has been applied. The equivalent current concept states
that the source of the diffracted field is ascribed to a fictitious equivalent current,
both electric and magnetic, flowing along the edge.
Consider the finite wedge, with excluded angle n and edge length L , directed in
the z direction as shown in Fig. 2.8. An arbitrary point along the edge can be chosen
as the origin, but it is convenient to set the origin at the center of the edge. The x
axis is directed along the normal to the edge lying on the front face of the edge, the
y axis coincides with the normal to this face. The incident wave is in the s ' direction
and the diffracted wave is in the £ direction. The angle between the front face and
the edge-fixed plane of incidence containing the vectors s ' and £ is 0'. The
corresponding angle for the edge-fixed plane of observation is <j>.
The incident field I?1 can be separated into TE and TM components (or parallel
and perpendicular components) with respect to the plane of incidence, i.e.,
£* = £&§' + £ i 0' = ^ + f i J w
(2.55)
The diffracted wave can also be separated into TE and TM components with respect
to the plane of observation, i.e.,
£?* = £ g 0 -l- £ 1 0 = £ j £ + £ 7^
(2.56)
Denote the equivalent electric current and equivalent magnetic current on the
edge as I and M, the scattered field will then be
sin0 j I (z)e]k2cos0 dz
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.57)
-3 1 -
J observation
T point
V
source
Fig. 2.8 Geometry of a finite wedge diffraction problem.
£ =
v
\ £o 4jc
- ■sin9
r
f M (z)c-,tecos0 dz
*
(2.58)
The expressions of I and M for arbitrary incident and observation angles have
been derived in [48]. For the special case of a plane conductor (i.e., n =2 ) and backscattering case (9 =
—0' <j>=
I and M can be simplified to [48]
Vm/ep (l+cos<|>)
M b ) = H ‘(z) - - - ■ ---- —
jk sin20cos<|>
Hz) =
(2.59)
_ H i( t ) z c o s e a - K ^ )
y/fcsin20cos<()
yit sin^sin(J)
where //z‘ and Ez‘ are the z components of the incident fields along the edge.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2 6 0 )
-32-
The back-scattered fields can be obtained by substituting Eq.(2.59)
and
Eq.(2.60) into (2.57) and (2.58) and are
(2.61)
E, =
Y
/COUo » -} *
47t
r
sin0 M (0) L
sin(fcL cos0)
kL cos0
(2.62)
The major difference between the finite wedge and the infinite wedge can be
summarized as follows:
1.
The diffracted fields of a finite wedge can be observed at any direction via the
s\n(kL cosQ)
term,
. If
then the diffracted field can be observed only
kLcosO
when 0 = 90°.
2.
For the infinite wedge, a TE incident wave can produce only a TE diffracted
wave and a TM incident wave can produce only a TM diffracted wave. That is,
the diffraction coefficient matrix is diagonal. However, the diffraction coefficient
matrix of the finite wedge is generally not diagonal. The equivalent current den­
sity depends on the z component of both the incident H field and the incident
E field as seen in Eq. (2.60).
Since the derivations of the equivalent current sources are based on the
geometric theory of diffraction, singularity occurs when the observation point is on
the shadow boundary or reflection boundary.
The back-scattered fields versus the elevation angle of a finite wedge with
length equal to 10 wavelengths, exterior angle equal to 2 ti , and incident azimuth
angle equal to 60° are plotted in Fig. 2.9. The back-scattered fields versus the
azimuth angle for L=10A., n=2%, and 0=85°, are plotted in Fig. 2.10.
2.3.4 Finite Plate
Exact expressions of the field scattered by a conducting finite plate do not exist,
because the boundary condition cannot be written as ux = constant, where ux is an
orthogonal curvilinear coordinate. The geometrical optics approximation cannot be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-33-
flQ
3
0
<u
a -15
Q.
-30
-45
20
55
90
125
160
125
160
angle in degree
e
u
• mm
1
- 15
I
-30
-45
20
55
90
angle in degree
(b)
Fig. 2.9 Plot of back diffraction fields versus elevation angle for n =2k , <j>=60°,
L = 10X (a), parallel polarization, (b). vertical polarization.
applied in the case of a plate. When the dimension of a plate is large compared to a
wavelength, both physical optics and the diffraction of a finite wedge as discussed in
the previous section can be used however to evaluate the RCS. In this section, both
methods will be described.
With the physical optics (PO) approximation, the induced current density is
Ts = I n x lt'
and the scattered field is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.63)
- 34-
horizontal
vertical
f
0
*3.
I
-W
-30
132
175
angle in degree
Fig. 2.10 Plot of back diffraction field versus azimuth angle for
L=10X, n =2n, 0=85°.
-y com-o e-j*
— [(nx/7‘ ) - (rixtf‘' /*)4 ] j
27160
plate
ds'
(2.64-a)
(2.64-b)
plate
where n is the unit normal vector of the plate, and (? is the factor preceding the
integral in Eq. (2.64-a).
An arbitrary polygonal plate with N edges lying on the z=0 plane is shown in
Fig. 2.11, where 7- is the vector from the origin to the z'th node point, and &,■ is the
unit vector directed from node i to node z+1.
The scalar scattered field E s is a two dimensional Fourier transform of the
shape of the plate. Define the function s(x,y) as
1 (x,y) inside the plate
0 (x,y) outside the plate
(2.65)
and
T - x'x + y'y
k (lr -/)) = k (sin0r cos<t>r -sin0,- cos0t-)Jc+k (sin0r sin<pr -sin0,- sin(j)t-)y
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.66)
- 35 -
> x
Fig. 2.11 An arbitrary polygonal plate with N edges lying on the z=0 plane.
+ k (cosBr-cosQi)z
= id + v y + w z = lc
(2.67)
Substitute Eqs.(2.65), (2.66), (2.67) into (2.64), it* can be expressed as
£* = ? JJ
dx dy = ^ 5 (u,v)
(2.68)
plate
The expression of S(u,v) can be proved to be [49]
l
S (u ,v) = — £ x &n ejtr * /„
j eiam**mdt
if v*0
(2.69)
0
JV n = 1
where /„ is the length of the nth edge, t is a temporary variable, Eq.(2.69) can be
further simplified as
-1
S (u ,v) = —
N
I
J V n=\
jk & n
«■<*„)■
jit-K
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.70)
-36-
S (u ,v) can also be expressed as follows if u*0
gJP-Z+i _ eJ?<
jJc &n
S(u,v)
= -7- X (yA»)’ inefi* .
JU n=1
The significance of Eq. (2.69)is: the scattered
if lc -&n * 0
if ?&„=<>
(2J1)
field of a conducting plate is
mathematicallyequivalent to the summation of the scattered fields of eachedge. The
contribution of each edge is proportional to -------- /„
v
sin(lnJc&n)
—------- . While the
lnk-&n
significance of Eq. (2.70) and (2.71) is: the scattered field of a conducting plate is
mathematically equivalent to the summation of the scattered field of each vertex. The
(*•& „)
->
contribution of each vertex is proportional to (— - — )/jk-6in.
The other approximation method is to consider the scattered field of a plate as a
summation of the diffraction fields of each edge. The contribution of each edge is
found by the previous section (Sec. 2.3.3). In this approximation, the extent of the
plate normal to the edge direction is assumed infinite. For each edge, one can specify
a set of edge-fixed coordinates for that edge and then find the incidence angle,
diffraction angle, TE and TM components of the incident wave and diffracted wave
in terms of the edge-fixed coordinates. Finally, one can use coordinate transformation
to convert the scattered field in the edge-fixed coordinate system to the laboratory
coordinate system. The procedures are shown in Fig. 2.12.
The difference of the back-scattered field of a plate using the two approaches
above has been numerically and experimentally compared
[57]. It was concluded
that the PO approximation method matches experimental results very well in the
proximity of main-lobe regions but deviates noticeably from these results at wide
angles from the specular direction. Outside the main-lobe regions, the second method
is more effective than the PO method. In a later chapter, we will use the second
method to numerically calculate the scattered field of a hexagonal plate over spectral
and angular windows of interest.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 37 -
f i x e d coord.
r ectan gular
coord.
4-
lo c a l coord.
s ph er ica l
coord.
r e ct an gu la r
coord.
lo c a l coord.
sp he r ic al
coord.
f i x e d coord.
re cta ngu la r
coord.
s p he r ic al
coord.
r e c t angular
coord.
I
s p he r ic al
coord.
■V
+
Fig. 2.12 Procedures of calculating the diffraction field of an arbitrary edge.
2.3.5 Corner Reflector
Comer reflectors or retro-reflectors are designed to provide a large RCS over a
wide range of aspect angles and frequencies. The most popular comer reflectors are
dihedral and trihedral comer reflectors, which are formed by the intersection of two
and three flat plates respectively. The angle required between the plates to achieve
the optimum cross section is a function of the bistatic angle. In the case of a dihedral
comer reflector, the required included angle is given by
|3 = 90°+a
(2.72)
where a is the bistatic angle.
The high frequency RCS patterns can be estimated by using geometrical optics
and taking into account multiple scattering. The approximate back-scattered cross
section of the dihedral comer reflector is given by [21]
167W2fc2sin2(jiV4 + b)
~
A*q
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 38 -
where a , b , and <(>are defined in Fig. 2.13.
Fig. 2.13 Geometry of a dihedral comer reflector.
The dominant contribution to the back-scattered field is due to the double­
bounce mechanism. Except for polarization either in the plane of incidence or per­
pendicular to the plane of incidence, the polarization of the scattered field will be
different from the polarization of the incident field due to multiple reflections.
2.4 Scattering Mechanism of a Complex Shaped Object
When the object becomes complex in shape, no exact analytic solutions are
available. The moment method, which uses the grid wire model, has been applied to
find the scattered field of complex shaped objects [50]. However, the computing time
is extensive especially when the object is large compared to a wavelength. Computer
simulation using the PO approximation has been conducted and SAR images of air­
planes were obtained [24]. However, the RCS of a complex object obtained by the
PO approximation method is valid only for certain range aspect angles because edge
diffractions and other scattering mechanisms are also major contributors to the scat­
tered field..
To a first order approximation, the object is separated into
a
number of
geometric shapes or segments, and the scattered fields from each shape are then
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-39-
coherentJy superimposed to form the total fields. However, care must be taken
because some shapes may be obscured by other shapes for certain aspect angles.
Thus the analytic components of the cross section of particular geometric shapes may
not be the same when the shape is in one part of the complex body as they are in an
isolated scatterer [21].
In the high frequency region, i.e., when the body dimensions are large in terms
of a radar wavelength, the scattered fields can be attributed to a combination of the
following components [21]:
1.
Specular scattering points.
2.
Scattering from surface discontinuities - edges, comers, tips etc.
3.
Scattering from surface derivative discontinuities.
4.
Creeping waves or shadow boundary scattering.
5.
Traveling wave scattering
6.
Multiple scattering points.
In the following, we will briefly discuss each of these scattering mechanisms.
2.4.1 Specular Points
The specular points for a given transmitter and receiver can be determined by
the ray optics, where the reflection law must be satisfied. The geometrical optics
cross section of specular points on a conducting surface is given by
o = 7tp!p2
(2.74)
whereplt and p2 are the principle radiiof curvature at thespecular points.
If the sur­
face atthe specular points can be described by z = f ( x , y ), and theincident wave is
in the -z direction, then a is given by [21]
0 2/ /ax2>02/
(2.75)
/dy2) - (d2f / d x d y )
It is noted that the cross section of a specular point is independent of frequency and
depends only on the curvature at that point.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-40-
Specular point reflections can also be obtained by the asymptotic solution of the
physical optics expressions. In Eq. (2.34A), as
k goes to infinity, by applying the sta­
tionary phase method [51], the result is
A k j t j r) = - ^ e ~ jkr E
(2.76)
where 7j' are the vectors such that
dwr-it)-r]
da'
Ir *i
=0
(2.77)
and
//2| n _ t \.rn
= -<'■ J |
J
da'2
(2.78)
J
where d Ida' is derivative with respect to the surface curvature. The points
corresponding to the solutions of Eq. (2.77) are called stationary points, equi-phase
points, specular points, or scattering centers. The absolute value of the term
h (Tj Ox^o
is called the scattering strength for that scattering center. It is seen
that the locations of the scattering centers depend on the directions of lr and /, as
well as the shape of the metallic surface. The above analysis illustrates that the
objects in the high frequency regime can be regarded as of discrete form, consisting
of point scattering centers.
Consider the back-scattering case. Orient the coordinate system so that the z
axis lies along the back-scattering direction and passes through a point on the surface
with its normal pointing in the z direction. Define the surface of the object as
z —f {x ,y) and the coordinates of the specular point as [0, 0, y (0,0)]. By applying the
stationary phase method to Eq. (2.39), one can obtain the following asymptotic scat­
tered field solution [21],
it’(7, =
^ » * [ r- 2 /(0 .0 )H v ) v ^
/?o
(2.79)
where \jr is either zero, re/4, or k/2 depending on whether the surface in the vicinity
of the specular point is convex, saddle or concave; pj and p2 are the maximum and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 41 -
minimum radii of curvature of the surface at the specular point.
The back-scattering cross section is then
Hs
O = 47tr2 l — I2 = 7lp!p2
(2.80)
which is identical to that predicted by geometrical optics.
2.4.2 Surface Discontinuities
When one of the radii of the surface curvature goes to zero, an edge is formed.
If both radii of the surface curvature go to zero, a tip is formed. Geometric optics
cannot predict the scattered field of edges and tips because the RCS would be zero as
expressed in Eq. (2.74). In the absence of specular reflections, contribution from edge
diffraction becomes more important. The contribution due to edge diffractions can be
estimated using GTD as described in the previous section.
The contribution from a finite wedge can be obtained by the equivalent current
method. The procedures of calculating the scattered field were shown in Fig. 2.13.
An edge can be a straight edge or a curved edge. The dependence of the backscattered field on frequency and size depends on the type of edge (straight or
curved) and angle of incidence (at or away from normal incidence).
2.4.3 Surface Derivative Discontinuities
As shown in Sec. 2.2 the back-scattered field using the physical optics approxi­
mation can also be expressed as
(2.41)
_ -jke-J* j
2 nr
(2.81)
where / is the integral in Eq. (2.41). The RCS is then
(2.82)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-42-
/ can also be expressed as [21]
N
j =i
av,
dns'
(-D fl expO‘2/:/. ) (------------------)
n~ ( 2 j k ) n
'
a/;+
a/;-
(2.83)
where lj* and ly are points at which discontinuities in the derivative of the projected
area exist. It is noted that the frequency dependence of the n th surface derivative
discontinuities to / is proportional to k~n as can be seen in Eq. (2.83). Therefore, the
higher the frequency is, the less the contribution of the surface derivative discon­
tinuity will be. The above equations will be used later to account for the interpreta­
tion of microwave diversity imaging.
2.4.4 Creeping Waves
Creeping waves are characterized by the wave striking at the shadow boundary
of a smooth, curved, perfectly conducting surface to propagate along the body surface
in the shadow region, and emerge at the opposite shadow boundary.
The concept of a creeping wave can be illustrated by Fig. 2.14. In this figure,
the incident wave illuminates the scatterer and is diffracted at the shadow boundary
at point A, at which the incident ray is tangent to the surface. At this point, a por­
tion of the energy propagates on the surface of the scatterer, shedding energy by radi­
ation as it progresses in a direction tangent to the surface at point B.
The creeping wave can be described by a launching coefficient at the point of
capture, by a diffraction coefficient at the point of radiation, by an attenuation factor
which accounts for the rate of radiation, and by a description of the path on the
scatterer traversed by the creeping wave [44],
The creeping wave loses energy in proportion to the distance traveled in terms
of wavelength behind the smooth surface. Consequently, as frequency increases, the
contribution of creeping waves lessens.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-43-
E1'
Fig. 2.14 Creeping wave concept of diffraction by a curved surface.
2.4 J Surface Traveling Waves
When long thin bodies, such as wires, ogives, and prolate spheroids are
illuminated near nose-on incidence, the back-scattered field exhibits a large peak if
there is a component of the incident electric field tangential to the surface and in the
plane of incidence. The back-scattered field are dominately contributed by the rear
of the body [21]. The body acts as a traveling wave antenna in the end-fire mode
excited by the incident field. This scattering contribution is called the surface travel­
ing wave. The analogy between the operation of end-fire antennas and the surface
wave contribution to the scattered field on long bodies has been extensively analyzed
in [52].
The back-scattered cross section of a long, thin body due to a traveling wave is
related to a reflection coefficient determined by the rear shape of the scatterer [21].
However, there is no analytical means available for estimating the reflection
coefficient. The traveling wave phenomenon of a thin wire can be clearly observed
by Fourier transforming the broad-band frequency response and will be discussed in
Chap.7.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-44-
2.4.6 Multiple Scattering
For concave bodies or a body with comers, the contribution of multiple scatter­
ing to the RCS becomes more important. To determine the position of the multiple
reflection points, ray optics can be used. The ray should satisfy Snell’s law, i.e., the
angle of incidence should be equal to the angle of reflection. Assume now that the
multiple reflection points have been found, then the scattered field due to a set of two
multiple reflection points may be approximated by
Ak) =
2 KRr
lcy S 2
F { ) x f i 2(k)ejk{(f'-f *)r{]
(2.84)
where
f f tf ) = h Z E g l
(2.85)
tjjc ) =
(2.86)
27C/C21
kySi
where /? 2i is the distance between the two scattering points, f 2i is the unit vector
connecting scattering point 1 to point 2,
and "f{ are the position vectors of the
two multiple scattering points, S 1 and S2 are the second derivatives at points 1 and 2
as defined in Eq.(2.78), and h ( f{ ) and h t f { ) are the normal vectors at positions
T { and r 2/, respectively.
2.5 Polarization Effect
Polarization dependence of a target is usually expressed in terms of the polariza­
tion scattering matrix, which relates the polarization of the incident electric field to
the polarization of the measured scattered field. Mathematically, the relationship
between the scattered fields and the incident field can be written by
“
•
El
E\
a ll a \2
II
E s2
d 2\
Oft
i
or
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-45 -
[E* ] = [S ] [E* ]
(2.88)
where the subscript 1 and 2 represent any two general orthogonal polarization com­
ponents, and dij elements are, in general, complex quantities. The matrix [ S ] is
called the scattering matrix.
The most commonly used orthogonal polarization sets are parallel and perpen­
dicular, right-hand circular and left-hand circular. If the transmitting antenna and
receiving antenna have the same polarization state, the measurement is called a co­
polarized system. On the other hand, if the transmitting antenna and receiving
antenna have orthogonal polarization, the measurement is called a cross-polarized
system. For a conducting reflector, a wave which is right circularly polarized
becomes left circularly polarized ( and vice versa) upon reflection. Therefore, the
co-polarized cases include: a parallel (or perpendicular) polarized transmitted field
and a parallel (or perpendicular) polarized received field, and a right-hand (left-hand)
circularly polarized transmitted field and a left-hand (right-hand) circularly polarized
received field. The cross-polarized cases include a parallel (or perpendicular) polar­
ized transmitted field and a perpendicular (or parallel) polarized received field, and a
right-hand (or left-hand) circularly polarized transmitted field and a left-hand (or
right-hand) circularly polarized received field [21]. It is noted that the definition of
cross-polarized terms for the circular polarization case is totally reversed in some
papers [53]. In this dissertation, we will adopt the definition defined in [21].
In general, for targets possessing an asymmetrical shape, the scattered field is
polarized differently from the incident field. This phenomenon is known as "depolari­
zation” or "cross-polarization". However, depolarization is difficult to define in the
bistatic case, because the orthogonality of two polarization vectors at two different
spatial positions has no unanimous definition.
Under the physical optics approximation, it has been shown in Sec. 2.2.2 that
the back-scattered field of a conducting object is polarized in the same direction as
that of the incident field. Therefore, physical optics predicts no depolarization. Multi­
ple reflections usually cause the state of polarization to change. The monostatic
diffraction coefficients, Dv and Dp , of an infinite edge have different values as
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-46-
shown in Fig. 2.5. If the incident wave has right-hand circular polarization, the co­
polarized (left-hand circularly polarized) and the cross-polarized (right-hand circularly
polarized) components of the diffraction coefficients versus the azimuth angle are
shown in Fig. 2.15. It is interesting to note that the cross-polarized component of the
diffraction coefficient is nearly constant over the whole aspect angles.
co
T3
3
TEL
-15
-30
135
180
angle in degree
Fig. 2.15 Diffraction coefficients of an infinite edge vesus azimuth angle. The
illuminating waves are circularly polarized, 0j =0^=90°, ty=$d ,n=2.
Diffraction at a finite edge will usually cause depolarization except when the
inclination angle 0 is equal to 90°. A nonzero <|> component can exist in the backscattered field when the incident field is in the 0 direction. This can be examined in
Eqs. (2.59) to (2.62), where both the equivalent electric and magnetic currents and
both the two orthogonal components, E q and E^, of the diffracted fields remain
nonzero in the equations when 0 is not equal to 90°. Furthermore, the ratio of the
two orthogonal components is a function of incident angles (0 and <|>) only and is
independent of the wedge length and frequency if the assumptions of the equivalent
current method hold. Assume both the transmitting and receiving antennas are circu­
larly polarized. The ratio of the co-polarized component to the cross-polarized com­
ponent of the back-scattered waves can be measured for various angles of incidence.
Conversely, if the ratio of the two components is measured, one can roughly estimate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-47 -
the incident angles with respect to the edge-fixed coordinates. In other words, the
polarization status of the back-scattered field provides feature information about the
wedge.
At high frequencies, as described in the previous section, the scattered field of a
complex object can be viewed as contributed to by discrete sources with different
scattering mechanisms. Usually, the major contributors are specular reflections, edge
diffractions, and multiple reflections. Accordingly, the scattered fields due to these
discrete sources can be expressed as
Z ( k ) = 2 3 (k)ej/f'r‘ + 2 tjihye***' + 2 ^m(k)ej f ^
i
j
m
(2-89)
where P = k (lr - lL), a) is the scattering vector of the ith scattering center, 5* is the
diffraction vector of the j th edge, and l?m is the scattering vector of the /n th multiple
reflection. Under the high frequency assumption, all 3}’s are independent of fre­
quency and have the same polarization as the incident wave; all by’s are independent
of frequency and their polarizations can be different from each other and different
from that of the incident wave; 7?m is independent of frequency but its polarization is
dependent on the ray paths traveled and is usually different from that of the incident
wave.
The advent of high resolution radar makes the expression of Eq. (2.89) very
valuable, because high resolution radar has the capability to separate each discrete
scattering center on the target. From the polarization status of each source, one can
estimate its scattering mechanism and its significant feature. Although the polariza­
tion of each source is independent of frequency, the polarization of Ps (k) can
changes with frequency. The polarization of J?s (k) is determined by the complex
ratio of the two orthogonal components of the total field. The ratio is a function of
frequency because of the terms e P 'r .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-48-
2.6 RCS Measurement
The RCS of a target is the equivalent area of the target assuming the target rera­
diated the incident power isotropically. It is a function of the target’s geometry, fre­
quency, aspect angle, and polarization. Mathematically, the RCS a is defined by
c(* ,04,0',40 = lim4itf2
(2.90)
The multiplication by 4nr2 is to make a independent of r.
Definition of the coordinate systems used are required to describe the RCS
measurement. When static scattering measurements are made, the illuminating
antenna is usually fixed and the object is rotated. A stationary coordinate system is
defined relative to a fixed reference point Body motions relative to the reference sys­
tem can then be described by three angles, the yaw angle, the pitch angle and the roll
angle. Scattering data for all directions of illumination can be obtained by fixing
either the pitch angle or the roll angle and by varying the remaining two angles [21].
Once the coordinate system is defined, the polarization of the antennas and the
scattering matrix can then be determined. The measured received signal Vs can be
expressed by
Vs = c [hr ] [ U r 1 [ S ] [ U ] [ ht ]
(2.91)
where [hr] and [ht \ are the polarization vectors of the receiving antenna and
transmitting antenna respectively, [S] is the scattering matrix as defined in Eq.(2.87),
[£/] =
cosy siny
-siny cosy
is the transform matrix with y the polarization angle (see
Ref. [21], p.896). If the same antenna is used for both transmitting and receiving,
then a 12 = a i\- if the target possesses a plane of symmetry, and the matrix elements
are referenced to this plane, then a 12 =^2i =0- The calculation of the amplitude and
phase of the elements of the scattering matrix has been discussed in [21].
Intrinsic to the definition of RCS is the assumption that the incident field at the
scatterer is a plane wave and the scattered field varies as 1/r. If the observation dis­
tance is long enough, the conditions above can usually be satisfied. The distance
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-49-
criteria are usually developed according to the phase and amplitude of field variation
along the transverse and longitudinal direction. The constraint that the transverse
phase variation between the incident field at the edge and center of the target extent
is within tc/8 results in the requirement that the distance Rm must be greater than
2L2/Xq, where L is the maximum dimension of the target as projected in a plane per­
pendicular to the direction of observation. In reality, the distance in the RCS meas­
urement conducted either indoor or outdoor sometimes cannot satisfy the required
conditions. To reduce the separation requirement and to obtain a nearly uniform
plane wave illumination, some "compact range" techniques have been proposed and
implemented [54], The fundamental concept of the compact range technique is usage
of a device which converts a spherical or cylindrical wave from the feed into a
quasi-plane wave. This conversion process is usually accomplished with the use of a
lens or a reflector [55].
If a full-scale target is not available for measurement or if the effect on RCS of
a design variation before the final design is chosen needs to be determined, building
a scale model of the target is useful. The theory of modeling EM systems has been
established [56]. Relations between the quantities in the full and scale systems are
found in [21], It is theoretically possible to model an EM system using scale factors
to make absolute measurements of parameters in the model system, and to relate the
measured quantities to the corresponding parameters in the full scale system. For
example, let the full-scale system be represented by unprimed quantities, and the
model system by primed quantities. Assume the permittivity and permeability of the
model medium is the same as those of the full-scale system. If the length / is scaled
by p times, i.e., I' = lip, the conductivity in the model system must be p times the
conductivity in the full-scale system, and the frequency be / ' = p f in order that the
two EM systems be equivalent. It is noted that the required far field distance can be
lessened p times for the scale model system, because
„ .
*-
2/ ';
-
x
2
-
21*
Vp
'
P\ -
p
<292)
Therefore, scale models are frequently used so that the far field condition can be met
by the measurement system.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-50-
The block diagram of the system used in out work for RCS measurement is
shown in Fig. 2.16 [10]. A test object is mounted on a rotating pedestal controlled
by a MINC-23 microcomputer. The transmitting antenna is a 36" parabolic reflector
fed by a Transco 9C 27400-2 righthand circularized spiral antenna. The return signal
is received by a dual polarization EM Systems A6100 2-18 GHz dual polarization
horn antenna. A Norsal 4115 90° hybrid combines the horn antenna outputs to meas­
ure the circularly polarized signal. The received signal is amplified by an Avantek
AWT-1803911 FET amplifier and then fed to an HP 8350 network analyzer. The
signal is generated by a HP 8340A synthesized sweeper, which covers frequency
from 10 MHz to 26.5 GHz, and is then amplified by a Varian VZM-6991K3 TWT
amplifier. The computer controls the synthesizer and collects amplitude and phase
data from the network analyzer through the IEEE-488 interface bus. Therefore the
whole measurement is fully automatically controlled by a micro-computer,
where a,b, and $ are defined in Fig. 2.13.
To obtain the absolute RCS of an object, an absolute calibration procedure has
to be used. In the environment of an anechoic chamber, the calibration procedures are
as follows [10]:
1.
In the absence of the object, measure the clutter fields, which include reflection
from the ground, absorber, supporting holder etc., and denote the clutter field as
N( f ) .
2.
Choose an object as the reference object, whose RCS must be known or can be
theoretically calculated. Denote the RCS as aR ( f ). The reference object can
be a sphere, a cylinder, or a comer reflector. Measure the field in the presence
of a reference object. Assume that the presence of the reference object won’t
disturb the clutter field, the measured field will then be SR( f ) = N ( f ) + R( f ) ,
where R ( f ) denotes the field scattered by the reference object only.
3.
Remove the reference object and replace it by the test object, the measured field
will then be Sa( f ) = N ( f ) + 0 ( f ) , where 0 ( f ) is the scattered field due to the
object only. The RCS of the object a0 ( f ) is then
o0( f ) = ~ ~ ^ ~ <5r ( f )
\R(f)\2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
COHERENT
RECEIVER
-T
I
RECEIVER
REFIRENCE
FREQUENCY
U)
SOURCE
TRANSMITTER
COMPUTER
PEDESTAL
CONTROL
Fig. 2.16 Block diagram of the RCS measurement system.
-52-
' S0( f ) - N ( f ) ,2
a93)
If a complex field measurement is desired, the phase relative to the reference
point (usually the rotation center of the target) is needed. In our measurement sys­
tem, a long metallic cylinder with its front rim placed to coincide with the rotation
center is found to be a good way to obtain a phase reference. However, the exact
RCS of a long cylinder is difficult to evaluate over a wide frequency range because
the beamwidth of the transmitting antenna is not constant and cannot cover the whole
cylinder. Therefore, the measured magnitude can not reflect the true RCS of the
cylinder. A conducting sphere with proper radius will be more suitable as a standard
reference object because the sphere can be within the main-lobe of the transmitting
antenna for each frequency. The magnitude of the measured field can then be com­
pared with the theoretical value, from which the frequency response of the measure­
ment system can be obtained. However, it is more difficult to calibrate the phase
reference by using a sphere because the center of the sphere is difficult to align with
the rotation center. In this situation, we recommend the following calibration pro­
cedure:
1.
Measure the clutter noise N( f ) .
2.
Mount a long cylinder with its front rim coincident at the rotation center. Meas­
ure the scattered field Sc ( f ).
3.
Replace the cylinder with a sphere with known radius. Denote the scattered field
as Ss ( f ) and the theoretical RCS of the sphere as os ( f ).
4.
Replace the sphere by the test object. The scattered field is St ( f ).
The complex scattered field of the object with respect to the rotation center is
then
" \Sc( f ) - N ( f )
Et ( f ) =
Ss ( f ) ~ N( f ) I ’ '
7)
(2-94)
and the absolute RCS of the target is
IE ,(H I2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(2.95)
-53-
Clutter is the reflection from all objects other than the test object itself. The above
analysis assumes that the background scattering is not affected by the presence of the
object. In some situations, the presence of the object did affect the background
scattering. For example, a tilted plate might reflect the incident wave to the holder
and then reflect back to the receiving antenna. In that case, coherent subtraction
won’t improve the results.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 3
IMAGE UNDERSTANDING IN
MICROWAVE DIVERSITY IMAGING
T he appearance of an image depends on the scattering properties of the object and
the imaging technique utilized. For example, the microwave scattering properties of
an object are different than those noted using optical waves. Optical wavelengths are
usually smaller than the surface irregularity of most objects. Optical illumination is
scattered diffusely from such rough surfaces yielding image information about the
whole surface. However, a surface with the same roughness can be smooth compared
to microwave wavelengths and the scattered waves may not be seen in most aspect
angles. Therefore, microwave images can be entirely different from their optical
counterparts. The imaging scheme used directly determines and affects image qual­
ity. For example, the image resolution obtained by conventional incoherent sidelooking radar is much poorer than that obtained by synthetic aperture radar [3].
Therefore, understanding and interpretation of an image should be based on both
knowledge of the scattering properties and the image formation technique imple­
mented.
Basic scattering theory of a perfectly conducting object has been reviewed in the
previous chapter. In this chapter, we will relate these scattering properties to the
reconstructed images obtained by a specific imaging scheme, microwave diversity
imaging, achieved by combining angular, spectral and polarization diversity. In this
imaging system, the object is seated on a rotating pedestal and the back-scattered
field is measured. A synthetic imaging aperture is achieved in this fashion costeffectively without resort to an array aperture. A very famous imaging technique
which also involves rotating the object is computer aided tomography (CAT), which
is concerned with the reconstruction of a section of the human body by means of Xray taken at different orientations. Basically CAT is a problem of image reconstruc­
tion from its projections. This problem has wide applications including medical
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-55-
imaging, geophysical signal processing, astronomy, and microwave imaging.
In this chapter, we will start by reviewing the image reconstruction of an object
consisting of point scatterers. We assume each scatterer scatters isotropically for
each aspect angle and no mutual coupling exists among them. This analysis is helpful
in illustrating the connection between reflection type microwave imaging and
transmission type CAT.
Determining the shape of a perfectly conducting object from a finite number of
measurements of its scattered fields is one of the purposes of microwave diversity
imaging. The PO approximation is usually applied to relate the geometrical shape of
a conducting convex body to its scattered field. In Sec. 3.2, we will summarize the
formulation of microwave imaging for conducting objects based on the PO approxi­
mation.
As we indicated before, the PO approximation is inadequate for the scattering
problem of a complex shaped metallic object The point-scatterer model is also not
suitable for describing the scattered field of a perfect conductor. However, we still
apply the reconstruction algorithm used in CAT to reconstruct microwave images.
Therefore in Sec. 3.3 we will give an alternate interpretation to the reconstructed
images based on the understanding of the scattering mechanisms and the reconstruc­
tion algorithm.
Several typical numerical and experimental examples will be demonstrated in
Sec. 3.4. These examples will illustrate the influence of the various scattering
mechanisms, angular windows, spectral windows, and polarizations on the recon­
structed image.
3.1 Image Reconstruction of Point Scatterers
Assume that the image of an object consisting of point scatterers is to be
obtained. Each point scatterer scatters isotropically for each aspect angle and there is
no mutual coupling among the scatterers. A plane wave with wave number k is used
to illuminate the object and the scattered wave is received. The object is then rotated
and the measurement is repeated. Let O be the reference point and the rotation center
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-56 -
of the object, z axis be the rotation axis, and a be the bistatic angle. The lines con­
necting the rotation center to the transmitter and receiver are both perpendicular to
the z-axis, that is, the antennas are in the x - y plane. The imaging geometry pro­
jected to the x - y plane is shown in Fig. 3.1. Let gt be the reflectivity of the ith
point scatter at (* ;,? /, z,-)» then the three-dimensional (3-D) reflectivity function of
the object becomes
u
x
y
V
r
receiver
- i
transmitter
Fig 3.1. Simplified imaging geometry projected to the X - Y plane.
g(x, y, z) = £ gt 8(x- x i,y - y t ,z-z{)
(3.1)
i
When the object is rotated through an angle <|>,the range and crossrange of the point
(x ,y) becomes u and v respectively. The relationbetween (u ,v) and (x ,y) is
u = xcos<t> + ysin<(>
(3.2a)
v = - x sin<(> + y cos<|)
(3.2b)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-57-
Therefore,
x = wcos(J) - vsin<t>
(3.3a)
y = Ksin<j> + vcos<|)
(3.3b)
The scalar scattered field with rotation angle 0 can be expressed as
£*(**) = J L f f f
**(U’V’Z) e-jW<n**'<n)
4jc£JJJ R*Sf)Rr( f )
_
jk
y g(|>(“ |»Vi»Zi ) ^-j^Rl+RD
4TO 7
(3.4)
/?//?/■
where R ‘(7*) and R r (F*) are the distances between the point T^and the transmitter
and receiver respectively, similarly, R f a n d R f are the distances between the ith
scatterer and the transmitter and receiver respectively. Under the far field approxima­
tion, the denominators R l (r*), R l (?) become
r
‘ (j *)= r ‘ -
(3.5)
R ‘ (ry) ~ R r
and the phase terms simplify to
R ‘ <f) + R r( D ~ R t + R r - 2kucos(oU2)
(3.6)
where R ‘ and R r are the distances between the rotation center and the transmitter
and receiver respectively. By substituting Eqs. (3.5) and (3.6) into (3.4), E s(k,$)
can be expressed by
£ a(*,0) = Jk6~m t +Rr ) 'L g ^ u i ,vi ,zi)ej2kUiCOs(aj2)
4jteR‘R r i
(3.7a)
= C( k) ^ g ^ u i ,vi ,zi )ej2kUiCOs(aJ2)
I
(3.7b)
Where C( k) denotes the term left to the summation in Eq. (3.7a). Let
8
1«)>(«, v ) = J g ^ u ,v ,z) dz
£ 2$(m) = / / g$(u,v,z) dzdv = j g ^(u,v) dv
(3.8)
(3.9)
whereg i0(w,v) is the projection of the object reflectivityfunctiononto the u -v
plane
at (u,v) and g 2q(u ) is the projection of thereflectivityfunction along the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-58 -
v-axis at range u. Eq. (3.7) can be rewritten as
£ 5(*,<|>) = C ( k ) j g 2<)(n )e '2tocos(a/2) du
(3.10)
If the term in front of the integration is removed, the range-corrected field E^ (k ,<)>)
becomes
Es' ( k S = jg^(u)ej2kucosandu
(3.11)
It is noted that E* (k,<)>) and g ^ i u ) are a Fourier transform pair. E* (k,<J>) is referred
to as the Fourier space data. If the Fourier space data are available over sufficient
bandwidth, i.e., over a sufficient range of wavenumber k , the projected data g ^ u )
can be estimated by a FT of the range-corrected fields or Fourier space data E* (fc,<j>).
There are some similarities and differences between CAT and microwave diver­
sity imaging systems. Both systems involve rotation of the object However, the
measured quantities in CAT are the projection data along the rays; while the meas­
ured quantities in the microwave diversity imaging systems are the Fourier space data
which are the FT of the projection data along constant ranges. If we inverse FT the
Fourier space data for each aspect angle in the microwave diversity imaging system,
we have the estimated projection data along constant ranges at each aspect angle. At
this point the two imaging systems are equivalent because both call for image recon­
struction from an object’s projections. The reconstruction algorithms used in CAT
can then be applied to this imaging system. Two methods have been used to recon­
struct the computer tomographic image: the direct Fourier method and the backprojection method. These two methods yield equivalent results, as proven by the
central-slice theorem or the projection slice theorem [27], and will be described
below.
Consider the two-dimensional case, let g'(x,y) be the density function to be
reconstructed from its projections. The FT of g' is defined as
G' ( X, Y) = J { g, (x,y)e-^xX+y^dxdy
so that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.12)
J
g' ( x, y) =
/ G 'Q C 'Y y e jW -W d X d r
—«o
(3.13)
oo
The projection of %’ along v -direction at angle <|>is given by
oo
P$(u)=
J
g'(ncos(|>-vsin(l>, nsin<(M-vcos<j))dv
(3.14)
The function p$(u) represents a series of such line integrals for each value of <|>. The
1-D FT of p$(u) is given by
oo
P*(U)= j p $(u)e~juUdu
oo
= J
oo
=
oo
j g'(ncos<|>-vsin<J>, Msin<(H-vcos<t))e-^ d n d v
■ oo
/ J
—o o
g' (X ,y )e-j(Uxcos^ uy**#>dxdy
oo
= G' (Ucos<(), U sin<{>>
(3.15)
The statement of the projection-slice theorem is
P ^ U ) = G '(U cos<J), U sin<(>),
(3.16)
that is, the FT of the projection at angle <|> is a "slice" of the 2-D transform G'(X ,Y)
taken at an angle <|>with respect to the X axis.
If the FTs of the projections (or the Fourier space data) at various angles <(> are
available, then the density function can be reconstructed by inverse FTs of the
Fourier space data. When a reconstruction is obtained by using this central-slice
theorem, the method is referred to as a direct Fourier method. In microwave fre­
quency and angular diversity system, the Fourier space data are usually in polar for­
mat. To make use of the fast algorithm of FT, one should convert the polar format
into rectangular format [58].
The FT relation in Eqs. (3.12) and (3.13) can also be written in polar coordinate
form. Let (r, \j/) and (p, q) be the polar coordinates of the space (x, y) and Fourier
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-60-
space ( X , Y ) respectively, the relationships between these pairs are
x = rcosy,
X = pcosrj,
y = r siny
Y = r sinrj
(3.17)
The polar expression of Eq. (3.13) is then
2k
oo
g '(r,y ) = J J pG '(p,ri)e-/2ro’pcos(11-v)rfpdTj
o o
2k
oo
= J
o
jpG ' (p,T\)ej2nrpcos<^ v)dp dr\
o
= j F [pG' (p,Ti)](r cos(T|-\|0)d q
o
(3.18)
where F [ ] is the FT of the bracketed quantity with respect to p. The above equa­
tion is the basis of the filtered back-projecdon algorithm. The procedures of the
algorithm are as follows:
1.
Fourier transform the product of the Fourier space data at each angle
<j> and
the
function p. The function p is also called the rho -filter [26]. This output is
called the filtered projection.
2.
The output is then uniformly back-projected in the v-direction, this backprojection line is called filtered back-projection line. Since the operation of
computing involves starting with a 1-D function to obtain a 2-D one, this opera­
tion is referred to as back-projection.
3.
The contributions to a pixel at (r ,\|/) from different filtered back-projection lines
are added coherently, i.e., polarity is taken into account.
It is noted that the multiplication of the Fourier space data by p will make the
point-spread function a delta point if the Fourier space data are available for all fre­
quencies and all directions.
Compare the above 2-D analysis to the microwave frequency diversity imaging,
we have the following dualities:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 61 -
g'(x,y)-+-+. gi(x,y)
U ■*—+- 2kcosal2
Therefore, the reconstructed image in the microwave diversity imaging is the
estimated reflectivity function g\(x,y), which is an image projected onto the z =0
plane.
In practice, the spectral band and the angular window of a microwave
im agin g
system are usually restricted by practical constraints. In that case, the point-spread
function will not be a point but is space-variant and angular window dependent. The
effect of the various angular windows on the point-spread function has been dis­
cussed in [12].
3.2 Formulation of Microwave Imaging for Conducting Objects Based on the PO
Approximation
Practically, a conducting object does not consist of discrete point scatterers. In
particular, a conducting object has obstruction effects, where various of its parts are
seen for certain viewing angle only. In addition, polarization is also an important
scattering factor. Therefore, the point-scatterer model is not suitable for a metallic
object because the scattering properties of the latter are dependent on frequency,
aspect angle, and polarization. In this section, we will investigate the formulation of
microwave imaging for a metallic object based on the PO approximation.
Under the PO approximation and far field condition, the back-scattered field of a
convex metallic object has been expressed in Eq. (2.38) and is repeated
(3.19)
•>ia
If we divide the object into two regions, lit region 1 and shadow region 2, and
express the boundary cross section as A , then by the divergence theorem, that is, the
closed surface integral of a vector B is equal to the volume integral of the divergence
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-62-
of the vector,
= jV&dv,
(3.20)
we can write Eq.(3.19) as
j
eJ™' % d ? f - j e j2kf' H r d?'
Sill + A
jke~jkr go
2nr
region 1
jke~jkr go
2nr
region 1
JJJ V ie j7ki' r lr )dv - fe i7jd'r lr d?'
jjj
A
j2Jcej7jd' r dv - je i2kir'r ir <&’
A
e 'j *
(3.21)
where
jjj
j2 keiW ’*‘ dv -
region 1
(3.22)
A
is called the normalized scattered field.
If the transmitter/receiver is placed opposite to /, direction or the object is
reversely placed, then the lit region and the shadow region will be interchanged, and
the conjugate of E/ ( k , - l r) becomes
Es (k, - lr ) =
2k
j j j - j2 k e jVd, r dv - j e j2Jc(-l' y r {-lr)-{-(&') (3.23)
region 2
A
By adding E/ ( k, l r) and E /* ( k ,-lr), one has
E /(k ,lr ) + E / \ k - l r) =
e ^ d V
(3.24)
Use the relation 2klr = P , the above equation can then be written in the following
scalar form
= j j j e ^ d 3/
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.25)
-63-
Define the characteristic function y o f the scatterer as
1
In
for all T*' e v
(3.26)
A lc P U /llA rP
and
Ti p) = -k[Es (P )+E ^-P)] / k 2
(3.27)
then Eq.(3.25) can be rewritten as
oo
T(P) =
JJJ
(3.28)
dv'
It is seen that T(P) and yfj*) are a 3-d Fourier transform pair. The above relation is
called the "Bojarski Identity". If V(P) can be measured for all P , i.e., all frequencies
and all directions /r , it is possible to reconstruct the geometric shape of the object
from the inverse FT of T(P), that is,
oo
7(0 =
JJJr {?)e-jp r d p
(3.29)
It is noted that the FT pair is between y(7*) and T(P) but not between yff') and
E ',(P ) or E's (-P ).
A
A
For the bistatic case, i.e., lt * -/r , define
P = k(lr - l t )
(3.30)
and define an object scattering function ^ (J*
jn f ) ^
as
- [(/z (?)xfi?)jr]ir for all r on Siu
A = jo
elsewhere
<3-31>
After removing the term in front of the integral, Eq.(3.19) can be rewritten as
PS\ P ) =
J
P)e~j P r ds'
(3.31)
Sill
It was stated that PS(P) and yCr*, P ) comprise a 3-D FT pair [10,11]. However,
Y'JP) is not only a function of
is also a function of P.
T4', but also a function of lr and Pi or I,, that is, it
Therefore, the 3-D FT relation is not exactly satisfied
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-64-
between Ps (P) and the geometric shape.
All the above relations are derived under the physical optics approximation.
However, the PO approximation suffers from some difficulties. First, the PO approxi­
mation is valid only for high frequency, while the inverse FT relation in Eq.(3.29)
requires integration over all frequencies, including the low frequency and resonant
regimes, for which the physical optics approximation is not valid. Second, Eq.(3.29)
requires integration over all frequencies and all aspect angles. In most practical situa­
tions, however, only a limited set of frequencies and angular windows are available.
In [59], a scattering information aperture function was defined as unity or zero for all
values of P for which T(P) is known or unknown respectively. An integral equation
for incomplete scattering information was developed and a regularized analytic closed
form for that integral equation was suggested. In the next section, we will use a
different approach to consider the effect of finite spectral and angular windows on the
images.
Furthermore, the PO approximation is inadequate for scattering problems of a
complex shaped object. The scattering mechanism of a complex shaped object has
been discussed in Sec. 2.4 The main contributors to the scattered field are specular
reflections, edge diffractions, multiple reflections, and creeping and traveling waves.
Although the scattering properties of a metallic object are different from those of an
object consisting of point scatterers, we will still use the same reconstruction algo­
rithms, i.e., the direct FT method and the back-projection method. However, we will
give an alternative interpretation of the images reconstructed by these methods.
3.3 Interpretation of the Reconstructed Images
In the following, we assume the lines connecting the reference point to the
transmitter and receiver are both perpendicular to the rotational axis. As mentioned in
Sec. 3.1, the direct FT method and the back-projection method yield equivalent
results. The back-projection method, however, provides more physical insight into the
image formation process of microwave diversity imaging. As stated in Sec. 3.1., the
two basic procedures in the back-projection algorithm are first to obtain filtered pro­
jected data and then to back-project the outcome. In the following, we will rename
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 65 -
the projected data along constant ranges as the range profile. Mathematically, the
range profile at a specific angle is defined as the FT of the frequency response of the
range-corrected scattered fields at that aspect angle. Because the scattered fields are
measured only over a finite bandwidth and there is a lack of low frequency data, the
range profile obtained by a FT of the passband data will be different from that
obtained by a FT of the infinite data. Besides, the range profile and the filtered range
profile, which is the FT of the product of the range-corrected scattered field and a
ramp weighting function, have similar shapes if the ratio of the highest frequency to
the lowest frequency is not much greater than 1 as will be elaborated upon below. In
the following we will first discuss the physical properties of the range profiles.
As mentioned in Sec. 2.4, to a first order approximation, the scattered field of a
complex object can be expressed as the linear superposition of the fields scattered
from discrete specular points (or scattering centers), fields diffracted from the "visi­
ble" edges, and fields scattered from the multiple reflection points, creeping waves
and traveling waves, etc.
The range-corrected scattered field contributed by the scattering centers or spec­
ular reflections can be written as
=
(3 -3 2 )
i
where
is the scattering strength of the i th specular reflection point and is propor­
tional to the surface curvature at that point. A FT of the above equation with respect
to frequency will give peaks at the differential ranges of the scattering centers with
amplitudes equal to their scattering strength.
Where an edge is the dominant contributor, the range-corrected scattered fields
due to the edge have been expressed in Eq. (2.61) and Eq. (2.62) and are repeated
here,
r
Eq
=
• o f / m r Sin(fct,COS0)
sine
1(0) L — - -------—
kL cosG
. jk L cos0 „ - jk L cos0
= s i n e I (0 )
L
-~
—n
2jkLcosO
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3 .3 3 )
-66-
E , = ^ s i n 8 M(0H Sin< f C° ^
v
kLcosQ
_jkL cos0_ _
,___________
-yAL cos0
= ^ s in 8 M (0 )te
(3.34)
where both / ( 0) and M(0) are functions of 6 and <(>, where 6 and $ are defined in the
edge-fixed coordinate system. The FT of Eq.(3.33) and Eq.(3.34) over finite
bandwidth will give two peaks, located at range about ±(L/2)cos0, which are at the
differential ranges of the end points of the edge and have an amplitude roughly pro­
portional to l/(£Lcos6).
Where multiple reflection points are the major contributors, the scattered field
due to multiple reflections expressed in Eq.(2.84) can be used. The FT of Eq. (2.84)
will give peaks at a range equal to the differential ray path length covering the multi­
ple reflection points.
The expressions of the scattered fields due to creeping and traveling waves are
more complicated and are dependent on the geometry of the object and measurement
arrangement. We do not attempt to discuss these conditions mathematically, but will
show their effect through use of examples.
From the above analysis, it can be concluded that locations of peaks or dom­
inant features of the range profiles correspond to the differential ranges of specular
points, the end points of visible edges, and the differential ray path length covering
the multiple reflection points.
The alternative expressions of the corrected back-scattered field using the PO
approximation have been shown in Eq. (2.81) and (2.83) and are repeated here,
£s _
y
* » ■
where
j
exp(j2lcl )(
y
S
a
e
w
i
r
^
0
,
x
_ j^A iL L )
a
r$ 35)
if-
and lj- are points at which discontinuities in the derivative of the projected
area A (/) exist. dA(l)/dl is also called range reflectivity by radar worker [24]. If
the bandwidth available is finite, and the ratio of the highest frequency to the lowest
frequency of the measurement system is not much larger than 1, then the denomina­
tor can be approximated by
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where k ^ , k max, and k0 are the minimum, maximum, and central wavenumber of
the measurement system respectively. Substituting Eq.(3.36) into Eq.(3.35), the FT
of Eq.(3.35) will give peaks at ranges of those points with nonvanishing discontinu­
ous derivatives. For example, assume the range reflectivity is shown in Fig. 3.2. The
scattered field, which is the Fourier transform of the range reflectivity, can first be
obtained by successive differentiation of the range reflectivity with respect to range,
and then by applying some fundamental theorems to find the FT.
From the fundamental Fourier transform theorems, we have the following FT
pairs:
F ( co)
/( /? )
/'(/? ) h - * . j v F ( © )
f " ( R) -« -* --(02 F ( g ) )
We also know the FT of S(R - R a ) is
By using the above relationship, we
can easily obtain the scattered field, which is
-jRi<a
P-}R&
p-jR*o
p-jR*<o
E(a)=a' ^j tor +ar—r
z- o r - a^~r- o r - a‘Hyo)
<3 - 3 7 >
where a 2 and a 3 are the slope of segment 2 and 3 respectively. If the ratio of the
highest frequency to the lowest frequency of the measurement system is not much
greater than 1, then the denominator can be approximated by
1
to
1
CO2
1
_ 1
0.5(comax + (amin)
co0
1
Q)2
(3.38)
(3.39)
By first substituting Eqs. (3.38) and (3.39) into (3.37) and then taking the FT of Eq.
(3.37), the estimated target reflectivity (or range profile) is then
/ ( / ? ) = [ a LS(/?-/?!) + a 2S(/?-/?2) - a 38(R-R3) - a & R - R J )
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 68 -
0
differentiation
a1
\
f”
!
i
»
i
i
*
«
i
•
i
i
i
i
i
i
i
i
i
1*
-a
$
differentiation
a2
>i
f
-a 3
Fig. 3.2 An example of target range reflectivity together with i
differentials.
successive
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where * represents convolution, AR is the range resolution of the measurement sys­
tem and is inversely proportional to the bandwidth available.
The simple example and analysis above show that the range profile estimated
from passband data accentuates discontinuities of the reflectivity function or object
scattering function. From this perspective, we can also conclude that the position of
the peaks of the range profiles correspond to the differential ranges of those points
with discontinuous derivatives in the range reflectivity.
All range profiles for all aspect angles can be represented as a sinogram. Sino­
gram representation has been used in CAT [26] and will be applied to represent the
range profiles of various aspect angles. A sinogram is a 2-D intensity varied display
with an abscissa of differential range along with the range profile, an ordinate of
aspect angle, and intensity or brightness proportional to the magnitude of the range
profile. For an isotropic point scatterer, the sinogram is a sinusoidal curve because
the peak of the range profile at aspect angle <|> is at a distance equal to rcos(0-<}>),
where r is the distance from the point scatterer to the rotation center, 6 is the angle
between the x-axis and the line connecting the point scatterer and the rotation center,
and <|> is the rotation angle which is the angle between the x-axis and u axis. The
peak of the range profile for each aspect angle has the same magnitude in the point
scatterer case. However, the sinogram of a conducting object has different proper­
ties. When the object is rotated or the receiver is moved, the scattering centers will
migrate to those points which satisfy Eq. (2.77). For example, the sinogram of a
conducting sphere with its center located at the rotation center is a straight line but
not a sinusoidal curve, because the symmetry of the sphere makes the range profiles
independent of aspect angle. The sinogram of an edge is also different from that of
two point scatterers each located at the ends of the edge. Both cases have sinograms
that consist of two sinusoidal curves, but the intensity of the curves in the former
sinogram span a very large dynamic range because the equivalent scattering strength
of the end points of an ecige is highly dependent on aspect angle as explained earlier
in Sec. 2.3.3; and scattering from an edge can contain a very strong single peak at
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-70-
aspect angle corresponding to broadside.
After the sinogram is obtained, a 2-D image can be produced by the technique
of back-projection as discussed below.
The implementation of the back-projection algorithm should satisfy the follow­
ing requirements [26,29]:
(1). The back-projection line should contribute to those pixels which it intersects and
to no others.
(2). The contribution of the back-projection line to a pixel must be proportional to
the scattering strength.
(3). The contributions tc a pixel from different filtered range profiles must add
"coherently" such that the bipolar nature of the filtered range profiles is
preserved.
If only the scattering centers are considered, after back-projection, the image
reconstructed can be interpreted as "collections of the projected scattering centers that
are visible over the angular window used in data acquisition." To predict what the
image will look like over the angular window, one can formulate the equations that
describe the target surface, use Eq.(2.77) and Eq.(2.78) to calculate the locations and
the scattering strengths of the scattering centers for each aspect angle, and then use
back-projection to reconstruct the images. If the surface of the object is smooth, and
the rotational angle increment is small, i.e., the angular spacing between adjacent
range profiles is small, then the resultant image will be "the projected outline of the
scattering centers of the object over the angular window", which usually reflects the
projective outline of the shape of the object.
The image of an edge of finite length L can also be interpreted through backprojection considerations. To simplify the discussion, only the back-scattering case is
considered. For a specific aspect angle, the positions of the peaks in the range profile
of an edge are located at the projections of the end points to the line of sight.
Amplitudes of the peaks are functions of 0 and <|> and are proportional to sine
( kL cos 0 ), where 0, 0 are defined in the edge-fixed coordinate system. After
back-projection, the contributions of this specific range profile to the reconstructed
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-71 image will be two parallel lines oriented in the image plane at the aspect angle. All
similar back-projected line pairs for various aspect angles will pass through one of
the end points, intensifying the brightness of the end points. However, when the
aspect 0 = 90° is within the angular window (i.e., the aspect at which the incident
wave is normal to the edge is contained within the angular window) , the backprojection due to this range profile will be a single bright line because Lcos0 =0, and
sine ( kL cos 6 ) = 1. The bright line coincides with the location of the edge, and
hence an edge will be well portrayed in the final image. The above discussions infer
that the microwave image of an edge will be two bright points whose locations
correspond to the end points of the edge if the normal aspect angle is not included in
the angular window; otherwise a line joining the two points and representing the edge
will appear in the image.
3.4 Examples
Some examples of reconstructed images and their interpretations are given next.
In the following numerical examples, the frequency coverage is from 6 GHz to 16
GHz.
(a). Effect of monostatic and bistatic imaging system
In the first example, we show the dependence of the image on the data acquisi­
tion geometry (monostatic and bistatic schemes). Consider an infinite metallic
cylinder with radius 20 cm. The cylinder’s axis is parallel to the rotational axis. The
transmitting and receiving antennas have identical linear polarization which are paral­
lel to the axis of the cylinder. The fields scattered by the cylinder are calculated
using the formula shown in Sec. 2.3.2. The reconstructed images are obtained by
rotating the target over 360° and by changing the bistatic angle between the
transmitter and receiver over 360°. The scattering center of a cylinder in the monos­
tatic case is at a point (actually a vertical line) on the surface where the unit normal
vector to the surface is parallel to the directive of the incident wave, while the
scattering center in the bistatic case is at a point (a vertical line) of the surface where
the unit normal vector parallel to the bisector of the angle between the transmitter
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-72 -
and receiver. The bisector is lying in the azimuthal plane. The scattering strength of
the scattering center for the bistatic case is proportional to cos (a/2), where a is the
bistatic angle [21], Over the specified angular window, projective image of the
scattering centers is a circle for the monostatic case and is a semicircle for the bis­
tatic case. The positions of the scattering centers on the cylinder resulting from object
rotation (or equivalently by rotating the transmitter/receiver) and from independent
receiver rotation while the transmitter is fixed and the sketches of the back-projection
for these two cases are shown in Figs. 3.3(a) and 3.3(b) respectively. The numerically
simulated images are shown in Figs. 3.3(c) and 3.3(d).
(b). Effect of Angular Window
In the second example, we show the effect of the angular window on the recon­
structed image. The geometry of a hexagonal plate shown in Fig. 3.4(a) is chosen for
this purpose. The inclination angle between the plate and the rotational axis is
6 = 85°. The polarizations of the transmitting and receiving antennas chosen are
right-hand circularly polarized and left-hand circularly polarized respectively, which
means that the image is formed from the co-polarized scattered field. It is assumed
that the scattered fields of the hexagonal plate are contributed by the six edges. The
field scattered from each edge is calculated using the equivalent current method as
described in Sec. 2.3.3.
Additionally, the procedures of the calculation were
described in Fig. 2.12. The sinogram over angular window from <J>= 0° to <j) = 180°
is shown in Fig. 3.4(b). In the sinogram display the bottom line represents the range
profile of the first aspect angle while the top line represents the range profile of the
last aspect angle. The sinograms are displayed in linear scale. The dynamic display
range of the sinogram has been properly chosen so that weak signals will not be
overridden by strong signal. It is noted that the sinogram contains several intensityvaried sinusoidal curves. The images reconstructed from angular window 1 ( <|> = 0°
to 180° ), window 2 ( 0 = 0° to 110° ), window 3 ( <p = 70° to 180° ), and window 4
( <|) = 0° to 87 °) are shown in Figs. 3.4(c) to (f) respectively. In window 1, each
edge of the plate becomes normal to the line of sight during rotation of the object
and therefore all the edges are visible in the image of Fig. 3.4(c). In window 2, edges
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-73ROTATKIG OBJECT
( o r r o t a t i n g T/R)
back p r o j e c t i o n
T/R
T/R
T/R
T/R
(a)
bistatic diversity
back p r o j e c t i o n
T
Fig. 3.3 Positions of the scattering centers of a cylinder in its projective image and
sketch of back-projection for the case of (a), rotating both the transmitter/receiver;
and (b). rotating the receiver only; over 360°. (c). the numerically reconstructed
image of case (a), and (d) is the numerically reconstructed image of case (b).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-74-
3 and 6 are never normal to the line of sight and therefore are not visible in Fig.
3.4(d), but the end points of the edges remain visible. The image of Figs. 3.4(e) and
(0 can be interpreted in a similar manner.
(c). Effect of Multiple Reflections and Bistatic Angles
In the third example we will show the effect of multiple scattering and the
effect of different bistatic angles in the data acquisition system on the images
obtained. An object consisting of two infinite separated parallel cylinders is shown
in Fig. 3.5(a). The radius of each cylinder is 4 cm, and the distance between the two
cylinders is 60 cm. Both the transmitting and receiving antennas are right-hand cir­
cularly polarized which means that the image is formed from the cross-polarized sys­
tem. The multiple scattering phenomenon is more pronounced for this polarization
case, because the specular reflection alters the state polarization from RHCP to LHCP
and vice versa and will be rejected by this antenna arrangement Experimental and
numerical results concerning multiple scattering of this structure have been discussed
in [13]. Series expansion solutions were used to calculate the scattered fields, and the
computing time necessary to generate the field data is very lengthy. However, under
the high frequency approximation, multiple reflection points for a given set of aspect
angles and bistatic angles can be found by ray optics. The details of the derivation
are shown in the Appendix A. When the object is rotated, the differential ray path
(relative to the round trip from the rotation center) will also be changed. The plots
of differential path length versus rotation angle <(> and the numerically reconstructed
images with bistatic angle a = 0° , a = 16°, and a = 40° are shown in Fig. 3.5(c),
(d) and (e) respectively. In the bistatic case, there are two sets of multiple reflection
points which satisfy the Snell’s reflection law as shown in the Appendix A. The
differential path lengths of these two sets are different and are also functions of the
bistatic angle, which explains the difference between these plots of differential ray
path and the reconstructed images. It is noted that multiple reflections have produced
artifact in the image that does not correspond to the physical shape of the object.
(d). Effect of Creeping Waves
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
:o
id)
is)
it)
Fig. 3.4 (a). Geometry of a hexagonal plate and the measurement arrangement, (b).
Sinogram of the hexagonal plate over angular window (0° to 180°). Image recon­
structed from different angular windows: (c). window 1 (0° , 180°), (d). window 2
(0°, 110°), (e). window 3 (70°, 180°), (f). window 5 (0°, 87°).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Transmitter
a
0
d
ruiif.i-' i "
6o
)
J0
*’ 0
r o t a t i o n angle
-k|
)
3
.5
r o t a t i o n angle
,0
*•5
■* 5
rocacion angle
Fig. 3.5 (a). A dual-cylinder object, and plots of the its numerically calculated
differential range of multiple reflection path versus rotation angle and simulated
images obtained by using different bistatic angles equal to (b). 0°, (c). 16°, (d). 40°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
>I
-77-
The theoretical scattered field of a conducting sphere has been expressed in Eq.
(2.45). The oscillatory nature in the RCS plot as shown in Fig. 2.6 shows the
existence of the creeping waves. In this example, we will examine the spectral
dependence of the creeping waves and its effect on the reconstructed images. As
stated earlier range profiles usually give more insight into the scattering mechanisms.
The location of the peaks in the range profile gives the information about the path
traveled by the incident wave during scattering and the magnitude of the peaks in the
range profile represents the average scattering strengths of that scattering mechanism
encountered by the wave over the specified spectral window. The damped interfer­
ence pattern of the RCS plot in Fig. 2.6 is an evidence that the contribution of creep­
ing wave decreases as frequency increases. This fact can also be examined by con­
sidering the range profiles obtained with different frequency windows. The range
profiles of a conducting sphere obtained from different frequency coverages are
shown in Fig. 3.6. The first peak of the range profile is at differential range -a from
the center, which corresponds to the specular reflection from the front point of the
sphere; while the second peak is at a distance of about na/2 from the center, which
corresponds to the effect of creeping waves which travel around the shadow part of
the sphere before emerging to travel in the direction of the receiver.
Examining the height and width of the lobes in the response obtained with
different spectral windows, one can have the following observations:
1.
The widths of the lobes corresponding to the lower frequency window are
broader than those corresponding to the higher frequency window.
2.
The height difference between the two peaks increases as the central frequency
of the spectral window increases.
It is noted that the above figures are plotted on a dB scale. In the microwave
regime, the surface curvature of targets of interest is usually large compared to the
radar wavelength. One can conclude that the contribution of the creeping wave to
the range profile is much smaller than that of the specular reflection and can usually
be neglected.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
S
ca
45
c
30
15
0
(c)
distance in terms o f radius
(a)
60
ip h t r e ; k a ^ l 3 3 t o 46
n x > 66 ; d l t p ( 6 6 .3 6 )
23
T3
C
5)
3
o.
45
30
E
a
15
15 ■
0
j # ik.
- 3 - 2 - 1 0
1
A
2
3
distance in terms o f radius
(b)
(d)
Fig. 3.6 Range profiles and reconstructed images of a conducting sphere obtained
by various spectral windows, (a), window 1 (ka =0.2 to 13.3) (b). window 2
(ka =26.7 to 40) (c). window 3 (ka =0.2 to 26.7) (d). window 4 (ka=13.3 to 40)
-78 -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
60
-79-
The reconstructed image of the conducting sphere formed using different spec­
tral windows are also shown in Fig. 3.6 for an angular window spanning from 0° to
360°. As shown in Example 1, the contribution of a constant range line in the sino­
gram to the image is a ring. Therefore, the expected image will be two-concentric
circles. The inner circle has radius a and the outer circle has radius na/2. However,
the outer circle can hardly be seen in the image because of the finite dynamic range
of display and film used. From this example, we reach the conclusion that the effect
of the creeping waves on the reconstructed image can usually be neglected.
(e). Images of a Complex Shaped Object
The above numerical results can be related to and used in the interpretation of
more complex shaped objects. An experimental example of the reconstructed image
of a complex object is given next The test object a metalized 1:100 scale model of a
B-52 aircraft with 79 cm wing span and 68 cm long fuselage was mounted on a
computer-controlled positioner situated in an anechoic chamber environment. 201
equal frequency steps covering the 6.1 to 17.5 GHz range were used to obtain the
frequency response of the object The transmitting and receiving antennas are both
right-hand circularly polarized. The sketch of the B-52 airplane and angular, sino­
gram, and reconstructed image of the test object using data collected in an angular
window of 90° extending from head-on to broadside in 128 views is shown in Fig.
3.7(a), (b) and (c) respectively. The intensity or brightness of each pixel is propor­
tional to the average scattering strength of that portion over the angular window. This
fact suggests that microwave diversity imaging can be applied to the study of radar
cross section (RCS) management, which will be discussed in the subsequent chapters.
Over this angular window, there exist certain aspects when the edges of the
right wing and right tail are perpendicular to the bisector of the transmitter and
receiver. However, there is no such aspect angle in the window used when the bisec­
tor is perpendicular to the edges of the left wing and tail. Therefore, the edges of the
right wing and right tail can be seen clearly in the reconstructed image, while only
the end points of the left wing and tail are visible in the image, as shown in Fig.
3.7(c).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(a)
it
f
I
(b)
(c)
w
Fig. 3.7 (a). Sketch of the B-52 airplane and the angular windows used. (b). Sino­
gram of the B-52 airplane over an angular window (0° to 90°). (c). Image recon­
structed over angular window (0° to 90°). (d). Image reconstructed over angular
window (18° to 90°).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 81 -
If the angular window is reduced from 18° to 90° as shown in Fig. 3.7(a), the
reconstructed image becomes as shown in Fig. 3.7(d). Over this reduced angular win­
dow, there is no specular reflection from the fuselage and engines. This explains the
difference between Figs. 3.7(c) and 3.7(d). From the above experimental examples,
we can conclude that the microwave images represent a projection image of the "visi­
ble scattering centers, visible edges when the angular window contains views where
the edges are perpendicular to the line of sight, or end points of the visible edges
when these edges are never perpendicular to the line of sight". Note for bistatic
imaging we can define the line of sight for data acquisition as being along the bisec­
tor of the angle a between the transmitter and receiver.
The above examples show the validity of image understanding and interpretation
as applied to microwave diversity images generated under different conditions and
serve to illustrate that the image is dependent on the data acquisition parameters
(monostatic, bistatic, polarization state, frequency range, and angular window) and
hence care must be taken for proper interpretation of the image.
3.5 Discussion
The analysis and experimentation described in this chapter is based on far field
assumption. If the far field condition is violated, the constant phase line can no
longer be approximated by a line perpendicular to the bisector of the angle between
the transmitter and receiver, and the Fourier transform relation between the scattered
field and object function as shown in Eq. (3.28) or Eq. (3.31) is no longer valid. To
overcome the difficulties caused by near field imaging, the matched filtering method
has been proposed in the acoustical multi-frequency imaging [16]. This concept has
been applied to imaging of rotating objects and near field images can be recon­
structed by this method [17]. However, the computing time is very time consuming
because no fast numerical algorithm can be applied to this problem. Near field
acoustic image has also been experimentally studied [60]. All the above discussion
is based on the point scatterer model. For a complex shaped conducting body, the
expressions of the near fields are more complicated. A thorough theoretical analysis
on the near field frequency diversity imaging of metallic objects has not been
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 82 -
reported and should prove to be a challenging research topic. This aspect of research
is particularly fruitful for microwave nondestructive evaluation of dielectric and com­
posite structures.
Interpretation of the microwave images conducted in this chapter also assumes
that the transmitter and receiver both lie in the azimuthal plane perpendicular to the
rotational axis. In this situation, the image is a tomographic and projective image.
However, if the bisector is not perpendicular to the rotational axis, the image of a
point not on the x - y plane will not be a point but an unfocused ring [12,61] if the
same reconstruction algorithms are applied. Mathematical analysis can be found in
[12]. However, this phenomenon can also be easily predicted from the sinogram
representation [12].
Successful interpretation and prediction of the microwave image are fundamen­
tal to research in several areas, including target identification, classification, and radar
cross section management. Its application to radar cross section management studies
will be further discussed in the subsequent chapters.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 83 -
CHAPTER 4
EXTRAPOLATION OF AVAILABLE DATA
INTO MISSING BANDS AND ITS
APPLICATION TO RADAR IMAGING
Fine resolution of object detail in the reconstructed image is necessary to locate hot
spots of an object over the prescribed spectral and angular windows employing
microwave diversity imaging. It is known that the range resolution of a frequency
diversity imaging system is inversely proportional to the bandwidth covered by the
measurement system. However, due to limitations of the measurement system or a
restriction of the bandwidth allocation, the observed data can lie in a narrow band or
multiple restricted spectral regions known as pass bands. Several methods of extra­
polating the measured data beyond the observed regions have been proposed and
tested [62,63,64] in an attempt to achieve complete information in the unrestricted
spectral range. In these methods a prior knowledge of the maximum dimension of
the object is assumed, and an iterative procedure is applied. The use of linear predic­
tion for the interpolation and extrapolation of missing data and data gaps has also
been reported recently [65].
To increase the resolution obtained from spectral data of such a limited extent,
techniques of nonlinear power spectrum estimation have been used with notable suc­
cess [66]. These include the maximum likelihood method (MLM) [67], the Pisarenko
method [68], the Prony’s method [69], multiple signal classification (MUSIC) [70],
autoregression (AR) [71], linear prediction (LP) [72], and the maximum entropy
method (MEM) [73]. The autoregressive model is a predictive, all-pole model. It
yields information about the time series outside the known interval and has applica­
tions in many diverse problems. The basic idea of the MEM is to choose the spec­
trum which corresponds to the most random or the most unpredictable time series
whose autocorrelation function agrees with a set of known values. This condition is
equivalent to an extrapolation of the autocorrelation coefficients of the available time
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
• 84-
series by maximizing the entropy of the process [73]. Linear prediction is used to
predict the next sample value in terms of the present and past values, and the linear
prediction parameters are found by minimizing the error power [74]. The basic philo­
sophies of AR, MEM, and LP are different. However, these three methods are
equivalent for the case of a stationary Gaussian process [66].
Most nonlinear spectrum estimation techniques are developed to process the data
in the time or frequency domain. However, there is an analogy between the timefrequency domain and the space-spatial-frequency domain. In microwave diversity
imaging, for a given aspect angle, the frequency response of the scattered fields
corresponds to a set of time domain data, while the square of the absolute value of
the range profile corresponds to the power spectrum.
The linear prediction method is especially suited for those cases when the spec­
tra are discrete. Under the high frequency condition, it has been shown in chap. 2
that the scattered fields of a complex target can be ascribed to a few discrete scatter­
ing centers and edge diffractions. Additionally, the locations of the scattering centers
and their scattering strengths are independent of the operating frequency for a given
transmitter/receiver pair. This is equivalent to saying that the spectra (or range
profiles) of the scatterer are also discrete. These phenomena provide the motivation
for application of the linear prediction method to microwave diversity imaging.
Although the spectra estimated by the MEM or AR can be very sharp and well
resolved, this may not be an advantage in a microwave imaging system. If the data
are not sampled densely enough in the spectral domain, the sharp well resolved com­
ponent may be missed, and the results may not faithfully reflect the actual spectral
amplitudes. As stated in chapter 3, image reconstruction from microwave diversity
imaging systems involves coherent superposition of the data in the spectra, or range
profiles, of the scatterer (obtained at different aspect angles), where these are
estimated from partial data available in segmented bands. If the estimated amplitudes
of the range profiles obtained by the MEM or AR depart from the desired values
because of undersampling, image degradation will result. Furthermore, we are
interested not only in the magnitude of the range profile, but also in the phase of the
range profiles as required in coherent superposition.
To overcome the dense
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 85 -
sampling requirement and retain the phase information of the range profiles, it may
be preferable to extrapolate the data available from the various pass bands into the
vacant bands before the spectra or range profiles are formed.
An intuitive method of extrapolating the data beyond the observed region is to
predict the exterior data by using the same parameters obtained by the linear predic­
tion model. One of the most popular approaches to linear prediction parameters esti­
mation with N data samples is the Burg algorithm [75]. For a given number of data
samples in a given observation interval, in order to separate the discrete spectra (in
this chapter, spectrum is defined as the Fourier transform of the observed data), the
required model order in the linear prediction method increases as the separation of
spectra decreases. In addition, for a given model order and given number of sampling
points, it is easier to distinguish the two close spectra components (scattering centers)
by a data set with longer observation interval than that by a data set with shorter
observation interval. It was also suggested that the model order should be constrained
to no more than half of the number of data points for short data segments so that
spurious peaks in the spectral estimate will be avoided [66]. From the above obser­
vations, one can conclude that it would be more difficult to resolve two closer point
targets with short data band than a longer data band. If all the observed data within
multiple restricted regions can be fully utilized, better resolution can be expected.
In the next section, a new iterative method which uses the Burg algorithm to
find the linear prediction parameters and then an iterative procedure to modify these
parameters is proposed and tested with both simulated and realistic measured data.
With this algorithm, one can obtain acceptable extrapolation beyond the observed
region if the spectra are in discrete forms and the separation of the spectra is not too
small. Both simulation and experimental results will be presented in Sec. 4.2 to
demonstrate the effectiveness of the method in microwave diversity radar imaging.
More discussions on this algorithm and others and their applicability to the extrapola­
tion of data from a finite angular aperture to an exterior angular aperture will be
given in Sec. 4.3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-864.1 The New Iterative Algorithm for Extrapolation of Data Available in Multiple
Restricted Regions
Assume the data sequence available is { X\..... xN }.If one wishes to predict xn
on the basis of the previous p samples, theforward linear predictor will have the
usual form
~ ~ ^ i^ p k ^ n -k
(4 .1 )
*=1
where apk is the prediction coefficient. Denote the forward prediction error for a
p th-order linear predictor as epn, then
ep n ~ * n ~
=
apkx n -k
k =0
(4 .2 )
where ap0 = 1.
One can also back predict xn_p on the basis of samples xn_p+l,...jcn, i.e.,
%n-p = ~ ^ i apk*n-p+ k
(4 .3 )
*=1
where * denotes the complex conjugate. Denote the backward prediction error for a
p-th-order linear predictor as bpn, then
bp n — * n - p ~ % n-P
~
( 4 .4 )
^Ld ^ p k ^ n - p + k
k=0
An approach to linear prediction parameter estimation was introduced by Burg [75].
The linear prediction parameters are obtained by minimizing the sum of the forward
and backward prediction error energies ep , defined by
A/-1
= 2
n =p
„
N-1
1V. 1 + S
n - p
1 bpn 1 ’
<4-5)
subject to the constraint that the prediction parameters satisfy the Levison recursion
relationship [66],
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 87 Qpp
—Q p-ijc
(4.6)
**ppt* p - l,p - k
for all orders from 1 to p.
If one is going to extrapolate from the available data beyond the observed
region, a straight forward way is to use the estimated prediction parameters apk and
the measured data in the following equations,
%N +j ~
*-j=
&pk * N +j - k
k=1
J ** 0 »
(4.7)
£ a %k x - j + k
k=1
J >0 ,
(4.8)
where Adenotes the estimated value.
If the data available are confined to multiple separate spectral regions or passbands of equal width as illustrated in Fig. 4.1, and one tries to extrapolate from the
observed data to the vacant bands, an intuitive method is to divide the inner vacant
band into two parts of equal width and to extrapolate into the left part by using the
prediction parameters obtained from the data set of region I and extrapolate into the
right part by using the model parameters obtained from the data set of region
region I
n.
^ region II
o
Fig. 4.1 Available data in multiple regions. Passband (shaded region) surrounded by
vacant bands.
If the data sequence can be correctly expressed by the prediction parameters,
then the extrapolation error, which is defined as the absolute value of the complex
difference between the actual values (either computer generated or measured values)
and extrapolated values, would be very small. However, if the prediction parameters
cannot model the sequence correctly, the error of extrapolation may accumulate. We
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 88 have found that the linear prediction model which characterizes the data sequence is
more accurate for longer data strings and larger model orders, especially in the pres­
ence of noise. However, the model order should not exceed half of the number of
the samples because the estimated spectrum will produce spurious peaks [66].
In order to utilize the information available in different regions, a new iterative
algorithm using the Burg algorithm to estimate the prediction parameters is proposed.
The procedure illustrated in Fig. 4.2 is as follows:
1.
Divide the inner vacant band into two parts of equal width. Extrapolate into the
left part by using the prediction parameters obtained from the data set of region
I and extrapolate into the right part by using the prediction parameters obtained
from the data set of region H If the bands are not equal in width, unequal divi­
sion of the vacant intervening bands may be appropriate.
2.
With the "vacant band'V data together with the observed data, use the Burg
algorithm to find a new set of prediction parameters.
3.
Using this set of prediction parameters and the data of region I, extrapolate into
the left part of vacant bands, and using the same set of prediction parameters
and the data in region II, extrapolate into the right part of the vacant bands.
4.
Using this set of parameters together with the extrapolated data, estimate the
data in the observation region I and II. Calculate the error energy between the
measured data and the estimated data in the observation regions. The error
energy is denoted by E j and is given by:
Ei = £
i
" * i |2 + '*/ - * 'i I2 = £ \ei \ 2 + \bi I2 ,
i
(4.9)
where xt are the measured data, £f are the forward estimation of jq , x \ are the
backward estimation of x, , e-t is the forward prediction error, and £>,• is the
backward prediction error.
5.
With the measured data together with the estimated "vacant band'V data, use
the Burg algorithm to find a new set of prediction parameters. From the meas­
ured data and this new set of prediction parameters, extrapolate the "vacant
band'V data as described in step 3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ITERATION ALGORITHM
1.
* AR PARAMETERS ESTIMATION Uj}i
--------------J---------------
Available data In regions I & II
are used eo extrapolate Into regions
III & IV.
Use data in regions I + II + III + IV
to estimate the parameters (aj}i,
where 1 represents the iteration
number.
Use data in I and faj >i to extrapolate
Into region III. Use data in II and
{aj}i to extrapolate into region IV.
1.
Convergence Test
a. Use data In III and {aj}l to estimate
new data values in region I. Use IV
and (3 4 ) 1 to estimate new data values
in region II.
* « •» # -1
II —
Calculate error
fii ’ EI + EII
For the resulcant data Is in atep 3:
If e^ < e j ^ , i+1 ■» i, go from step
3 to 2 otherwise iteration stopped.
Fig. 4.2 Schematic diagram of the new iterative extrapolation method proposed.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-90-
6.
Use the same procedure of step 4 to calculate the new error energy of the
passbands, call it E 2.
7.
Compare E \ with E 2, if E 2 is smaller than E t, replace the error energy Ej by
E 2 , repeat step 5.
8.
If E 2 is greater than E x, stop the iteration, and take the extrapolated data of the
previous loop as the final result.
In step 1, if the width of a single band (band I and/or band II) is not large
enough, the extrapolation errors produced by the prediction parameters obtained from
single passband data may be very large, in that case, we can set the data in the
vacant bands to zero.
The iterative method above can be easily applied to the case where only one
single data band is available. The procedures are almost the same except that only
one data band sequence is used to extrapolate to the exterior bands and to calculate
the extrapolation errors.
4.2. RESULTS
In this section, the performance of the new algorithm proposed will be evaluated
using both simulated and realistic data.
4.2.1 Range Profiles
First, assume for simplicity an object consisting of n point scatterers located at
( rQ + yj ) is illuminated by a plane wave, where r0 is the distance between the
transmitter/receiver and a reference point of the object and yj
is the differential
range of the j th scatterer (range relative to ra). Under the far field condition and
ignoring multiple scattering, and considering for simplicity a scalarized field, the
range-corrected scalar field can be expressed as
E' s (k) = ' Z a j e jllcy>
j
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(4.10)
- 91 -
In the following simulation, the theoretical values of E's (k) are calculated in
200 equally spaced frequency steps covering the frequency range f { = 6 GHz to
/200 = 16 GHz, with signal to noise ratio set to 40 dB. These values anticipate the
realistic experimental data utilized in testing the algorithm.
Assume the available (computed) data are in the following pass-band
(/3 0
>/ so )
( /
120
i/
170
)• We want t0 extrapolate the data to the vacant bands
( / l » / 2 9 )• ( / s i >f 119 )> and ( / 171 , / 2oo )• The range resolution obtained by the
DFT method using the whole bandwidth
( / 1 ./2 0 0
) *s about 1.5 cm. The resolution
using a single frequency band is about 5.5 cm. The resolution using both frequency
bands is about 2.0 cm, however, very high side-lobe level will be produced. We con­
sider a scattering object with seven point scatterers, the location and scattering
strength for each point scatterer are ( r j = - 30 cm, ay = 0.5), ( r 2 = - 20 cm,
a 2 = 0.5), ( r 3 = - 10 cm, a 3 = 0.5), ( r 4 = - 2 cm, a 4 = 1), ( r s = 10 cm,
a 5 = 0.25), ( r 6 = 20 cm, a 6 = 0.25), ( r 7 = - 30 cm, a 7 = 0.25). The values of the
field at each sampled frequency f j are calculated using eq. (4.10).
Define the extrapolation error at frequency f j as
£ ( / , ) = I E's ( f j ) - E ' s ( f j ) I ,
(4.11)
where E' s( f j ) is the extrapolated value at each f j . The extrapolation errors for
different algorithms are compared and shown in Fig. 4.3(a). The bold solid curve is
the amplitude of the theoretically computed fields E's ( f j ), the thin solid curves are
the extrapolation error after 100 iterations using the algorithm proposed in [63], the
dashed curves are obtained by using the Burg algorithm to find the prediction param­
eters from the respective pass-band, and using this set of parameters together with
data in each pass-band to extrapolate to the outside regions (bands HI and IV). The
dotted line curves are obtained using this new algorithm with one iteration and with
model order 25. The algorithm proposed in [63] involves basically application of the
Gerchberg algorithm to data in the multiple restricted regions. However, no numerical
or experimental results are given in that paper. It is clear from the results obtained
here that the algorithm in [63] seems to be ineffective in the case considered as the
errors can exceed the amplitude of the theoretical fields. The extrapolation obtained
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-92-
1
30
80
120
170
200
SAMPLE POINT
(o)
1.00
Ul
a
3
t
Z
Si
s
0.25 *
75 -75
?5
O
(c)
75
0
75
Id
a
3
H
Z
o
<
2
-75
(d)
DISTANCE (cm)
(e)
Fig. 4.3 (a) Magnitude of theoretical fields and comparison of extrapolation
errors of different methods, / 1 = 6 GHz, / 2oo = 16 GHz.
===== magnitude of theoretical fields.
extrapolation error from a single passband, no iteration.
extrapolation errors from new iterative algorithm.
extrapolation errors from algorithm proposed in [3].
(b) FFT of the whole band data.
(c) FFT of the passband data.
(d) FFT of the passband and extrapolated datawith 1 iteration.
(e) FFT of the passband and extrapolated datausingalgorithm proposed in [63].
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-93 -
fro m a sin g le p a ss-b a n d are n o t g o o d in th is e x am p le, b ecau se the m o d el o rd e r is not
su fficien t to m o d el th e d a ta se ries in th e p re sen c e o f n oise. T h e n e w m eth o d pro­
p o se d is se e n to p ro d u c e sm all e rro r a fte r o n e iteration .
T h e F o u rie r tra n sfo rm s (F T ) o f th e a ll-b an d d a ta (i.e, d a ta in reg io n I to IV ),
p a ss-b a n d d a ta o n ly , p a ss-b a n d p lu s e x tra p o la ted d a ta u sin g th e a lg o rith m in [63] and
p a ss-b a n d p lu s e x tra p o la te d d a ta w ith th e n e w m eth o d p ro p o se d a re sh o w n in Figs.
4 .3 (b ) to 4 .3 (e ). T h e se tran sfo rm s y ield th e ra n g e p ro file o f th e scatterin g o b je c t It
is c le a r th a t a F T u sin g p a ss-b a n d d a ta o n ly (F ig. 4 .3 (c )) h as v e ry h ig h sid e-lo b e
stru ctu re, th e F T o f th e e x tra p o la te d d a ta u sin g th e a lg o rith m in [3] (F ig. 4 .3 (e)) is
to ta lly d iffe re n t fro m th e o rig in a l o f F ig . 4 .3 (b ). T h e re su lt o b ta in e d b y F o u rier
tra n sfo rm in g th e d a ta g e n e ra te d b y th e p ro p o se d a lg o rith m is sh o w n in F ig . 4 .3 (d ). It
e x h ib its e x c e lle n t a g re e m e n t w ith th e re su lt o f F ig. 4 .3 (b ). T h e m ag n itu d es o f the
p e a k s in F ig s. 4 .3 (b ) a n d 4 .3 (d ) d e p a rt fro m the o rig in a l a ssig n e d v a lu e s b ecau se o f
th e z ero p a d d in g u se d in th e fa st F o u rie r tran sfo rm (F F T ) a lg o rith m . T h is lac k o f
fid elity in sc a tterin g stre n g th re co n stru c tio n d o e s n o t h a v e a d isc e rn ib le d eg rad in g
e ffe c t on th e q u a lity o f th e im a g e re c o n stru c te d as w ill b e illu stra te d b elo w , b u t is
im p o rta n t a n d m u st b e d e a lt w ith w h en a q u a n titativ e a n aly sis o f sc a tterin g stren g th s
is n eed ed .
th e fre q u e n c y c o v e ra g e is in cre ased to ( / i = 6 GHz, / 200 = 20 GHz) w ith
the n u m b er o f sa m p lin g p o in ts b e in g fix ed a t 2 0 0 an d the p a ss-b a n d s are k e p t at
( / 3 0 » / so ) a n d ( / 1 2 0 - f i 7o)>
c o m p u ted fields a n d the e x tra p o la tio n e rro rs w ill
be as sh o w n in F ig. 4.4 . It is seen th at the ex tra p o la tio n e rro r in d ic ate d by the d ash ed
lin e b eco m es sm a lle r d u e to th e in cre ased freq u en cy c o v erag e. I f th e freq u en cy c o v ­
e rag e is d e c re a se d to ( f \ = 6 G H z, /2 0 0 = 12 G H z), th e re su lts w ill be a s sh o w n in
F ig. 4.5. It is seen th a t th e e x tra p o la tio n e rro rs in d ic ate d by th e d a sh e d an d do tted
cu rv e s are n o w b o th h ig h . T h e F F T s o f the w h o le b an d d ata, th e p a ss-b a n d d a ta only,
an d th e e x tra p o la te d p lu s p a ss-b a n d d a ta u sin g this m eth o d are sh o w n in F ig s. 4.5(b)
to 4 .5 (d ) re sp ec tiv e ly . T h e re su lts in F igs. 4 .4 an d 4 .5 in d icate th e d esirab ility o f
u sin g se g m en ted sp ectral d a ta sp a n n in g w id e r sp ectral ran g es.
If
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-94-
SAMPLE POINT
Fig. 4.4 Magnitude of theoretical fields and comparison of extrapolation
errors with and without iteration, / 1 = 6 GHz, / 200 = 20 GHz.
— — magnitude of theoretical fields.
extrapolation error from respective passband, no iteration.
...............extrapolation errors from new iterative algorithm.
Although the above algorithm is an iterative one, it was found that extrapolation
errors usually decrease significantly after the first iteration, and further iterations do
not seem to improve the results. Therefore, it is practical and usually sufficient to use
only one iteration.
The performance of the algorithm using realistic data is also evaluated. The test
object, a metalized 1:100 scale model of a B-52 aircraft with a 79 cm wing span and
68 cm long fuselage was mounted on a computer-controlled elevation-over-azimuth
positioner situated in an anechoic chamber environment. 201 equal frequency steps
covering the / t = 6.1 to / 2oi = 17.5 GHz range were used to obtain the frequency
response of the object. The target is positioned for a fixed elevation angle of 30°
while the azimuth angle was altered between 0° and 90° in steps of 0.7° for a total
of 128 angular aspect views.
The pass-bands are first defined as ( / 30 , / 80) and ( / 120 >/ no)- The measured
values and the extrapolated errors of the broad-side perspective which is 90° from
the head-on orientation are shown in Fig. 4.6(a). The solid line curve is the amplitude
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-95-
MAGNITUDE
4
3
2
1
0
80
120
SAMPLE POINT
200
170
(a)
MAGNITUDE
1.00
0.75
1
-
-
0.50
0.25
-
j Il
0.00
-75
0
(b)
A
75 -75
0
75 -75
DISTANCE (cm)
(c)
0
75
(<3)
Fig. 4.5 (a) Magnitude of theoretical fields and comparison of extrapolation
errors with and without iteration, f \ = 6 GHz, / 2oo = 12 GHz.
—
-magnitude of theoretical fields.
extrapolation error from respective passband, no iteration.
extrapolation errors from new iterative algorithm.
(b) FFT of the whole band data.
(c) FFT of the passband data.
(d) FFT of the passband and extrapolated data with 1 iteration.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
-96-
of the range-phase corrected field. The dashed curve represents the extrapolation error
resulting from extrapolating from each single band (bands I, II) with model order 25
as described in step 1 of the proposed algorithm. The dotted line curves are obtained
using the new algorithm with 1 iteration and model order 25. The extrapolation error
for measurement is defined in a manner similar to the definition of error in the
numerical simulation as the magnitude of the difference between the corrected meas­
ured fields and the extrapolated fields. The Fourier transforms from the whole band
data, the pass-band data only, and the pass-band together with extrapolated data are
shown in Figs. 4.6(b), 4.6(c) and 4.6(d) respectively. A Fourier transform of the
corrected scattered fields will give the range profile of the target in that view. In this
figure, it is seen that the extrapolation errors do not improve after one iteration. The
reason stems from the plot of the range profile shown in Fig. 4.6(b). In this view
direction, the major contributions to the scattered fields are due to the fuselage and
primarily those engines and fuel tanks which are on the illuminated side. Specular
scattering from these points are well separated in time or distance and their number is
small. Hence the linear prediction parameters obtained from a single pass-band are
sufficient to model the data sequence. The extrapolation errors are not as small as
those obtained by simulations. The reason for this is that the applicability of the
linear prediction model to the extrapolation of scattered fields of a metallic object is
based on the high frequency approximation. In the measurement data, however,
polarization effects, edge diffraction, multiple scattering and the failure to satisfy the
high frequency approximation in the lower region of the frequency band utilized in
the measurement will degrade the performance of the algorithm.
For the view line #64, which is 45° away from broadside, the measured fields,
the extrapolation errors, and the range profiles reconstructed from the whole band
data, passband data only, and (passband + extrapolation data ) after one iteration are
shown in Fig. 4.7. It is seen that there are many peaks in the range profile and the
separation between the the peaks is small. At this aspect, no strong specular points
exist.
The major contributions to the scattered fields are from the weak edge
diffractions. Furthermore, no edge is normal to the bisector at this aspect. The
scattering strength due to the other mechanisms is usually a function of frequency. It
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
MAGNITUDE
- 97 -
SAMPLE POINT
(o)
MAGNITUDE
too
0.75
0.50
0.25
0.00
-45
45 -45
0
45 -45
45
DISTANCE ( c m )
( b)
(c)
( d)
Fig. 4.6 Measured Fields, Extrapolation Errors, and Range profiles of view line #1.
(a) Magnitude of the measured fields and comparison of extrapolation
errors without and with 1 iteration
—
magnitude of theoretical fields.
extrapolation error from respective passband, no iteration.
extrapolation errors from new iterative algorithm.
(b) FFT of the whole band data.
(c) FFT of the passband data.
(d) FFT of the passband and extrapolated data with 1 iteration.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
■ ,t.
6.0
J ,
8.9
i_________________
11.8
14.6
17.5
frequency in G H z
(a)
-65
0
distance in c m
65
-65
(b)
0
65
(c ) d istan ce in c m
a.
-15
-65
-65
distance in cm
(d)
(e ) distance in cm
Fig. 4.7 Measured Fields, extrapolation errors and range profiles of view line #64.
magnitude of theoretical fields.
extrapolation error from respective passband, no iteration.
extrapolation errors from new iterative algorithm.
(a) magnitude of measured fields and comparison of extrapolation errors
without and with 1 iteration.
(b) FFT of the whole band data
(c) FFT of the passband data
(d) FFT of the passband and extrapolation data with 1 iteration.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
-9 9 -
is known that the linear prediction model or equivalently, the AR model, is an all­
pole model, which assumes that the observed fields are contributed by discrete
sources, and their scattering properties (positions and scattering strength) are indepen­
dent of frequency. However, the physical realities violate the fundamental assump­
tions of the model at this aspect. Therefore, small extrapolation errors are not
expected from this aspect.
Shown in Fig. 4.8 are the measured fields, the extrapolation errors and the range
profiles of the view line #85. The front wing is normal to the bisection line in this
aspect The discrepancy between the range profiles estimated from the whole-band
data (Fig. 4.8(b)) and from the (passband + extrapolated data) (Fig. 4.8(d)) is small,
illustrating the effectiveness of the algorithm in this set of data.
4.2.2 Reconstructed Images
The reconstructed images of the test object using data collected in an angular
window of 90° extending from head-on to broadside in 128 aspects and different fre­
quency bands are shown in Fig. 4.9. A co-polarized transmitting/receiving system
was used. Figure 4.9(a) is obtained by using the whole band data; Fig. 4.9(b) is
obtained by using the pass-band data alone. Figures 4.9(c) and 4.9(d) are obtained
by extrapolating without iteration and after one iteration respectively. The model
order used is M = 25 in both cases.
If the pass-band is defined as ( / 65 , / 130), the reconstructed images obtained by
using the pass-band data alone and by extrapolation without iteration and after 1
iteration are shown in Figs. 4.9(e), (f), and (g) respectively. The model order used is
also M = 25.
It is seen that the image resolution of Figs. 4.9(c), (d) and Figs. 4.9(f), (g) has
been improved a great deal compared with that of Fig. 4.9(b) and Fig. 4.9(e). The
ringing phenomenon appearing in Fig. 4.9(b) is due to the use of multiple restricted
bands. For a filled antenna array, the angular resolution of the array is approximately
proportional to the total length of the array. If the array is thinned or is empty in
some portions, the resolution is still proportional to the total length of the physical
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 100-
15
"3
3
10
5
0
6.0
8.9
14.6
11.8
17.5
frequency in G H z
(a)
25
CO
•3
C
u
•a
3
0
distance in cm
a■a
65
0>)
-65
0
(c )
distance in cm
25
15
■ua
_______ 1
Q.
E
3
-5
-15
distance in cm
•65
i ih ii
(e )
0
65
distance in cm
Fig. 4.8 Measured Fields, extrapolation errors and range profiles of view line #85.
■■1 magnitude of theoretical fields.
extrapolation error from respective passband, no iteration.
extrapolation errors from new iterative algorithm.
(a) magnitude of measured fields and comparison of extrapolation errors
and with 1 iteration.
(b) FFT of the whole band data
(c) FFT of the passband data
(d) FFT of the passband and extrapolation data with 1 iteration.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
<e>
(f)
(g )
Fig. 4.9 R e co n stru c te d im ag es o f th e m etalize d scale m o d el B -52 a ircraft u sin g
an a n g u la r w fn d o w o f 9 0 ° e x te n d in g fro m h ead -o n to b ro ad -sid e in 128
lo o k s a n d d iffere n t sp ectral co v erag e.
R eco n stru ctio n s from :
(a) e n tire b a n d w id th i f x , f z o x ) (b) p a ssb a n d ( / 30 > / so)» ( / 120 • / no)(c) p a ssb a n d ( / 3 0 . / so)> ( / 120 • / no)
a n d ex tra p o la tio n d ata (ex trap o lated d a ta in to em p ty ban d s) w ith o u t iteratio n .
(d) p a ssb a n d ( / 30 , / 80), ( / 120 , / n o )
a n d e x tra p o la tio n d a ta w ith 1 iteration .
(e) p a ssb a n d ( / 65 , f l30).
(f) p a ssb a n d ( f 65 , / 130) and ex trap o latio n d a ta w ith o u t iteration .
(g) p a ssb a n d ( / 65 , / 130) an d ex trap o latio n d a ta w ith 1 iteratio n .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 102-
array, but high sidelobe levels or grating lobes will appear.
The image reconstructed from the passband (/65»/i3o) shown in Fig. 4.9(e)
looks fuzzy due to the insufficient bandwidth. The reconstructed images without and
with 1 iteration as shown in Fig. 4.9(f) and Fig. 4.9(g) look almost the same. This
fact indicates that extrapolation from a single band is sufficient to give good images
for microwave imaging applications.
The above results are obtained using the co-polarized data. When the same
extrapolation algorithm is applied to the cross-polarized data, reconstructed images as
shown in Fig. 4.10 are produced. These images illustrate the effectiveness of apply­
ing the extrapolation method to radar imaging regardless of the polarization being
used.
Fig. 4.10 Reconstructed images of the metallic scale model of a B-52 aircraft
using cross-polarized waves.
(a) Reconstructed from wholeband.
(b) Reconstructed from passbands (f 30, / so)
( f 120 / 170)
extrapolation
data without iteration.
4.3 Discussion
Three parameters can be varied in the iterative extrapolation algorithm. They
are: the model order number used in the extrapolation from a single band, the
number of iteration, and the model order number used in the iteration procedures.
The selection of a suitable combination of these three parameters poses an interesting
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 103-
problem as it is difficult to have a rule of thumb. Relevant factors include the pro­
perty of the scattering source, the width of the passband, and the signal to noise ratio.
If the number of discrete sources is large, separation between sources is small, or the
SNR ratio is low, then the model order number should be large. If the number of
discrete source is small and the separations are far apart, then small model order is
sufficient to model the data sequence. A criterion for the error energy has been
tested in the iterative algorithm. It was found that the iterations stopped upon an
increase in error energy. It was also found that extrapolation errors may not decrease
even though the error energy does, and the speed of convergence may be slow.
Therefore, it is better to set a maximum number of iterations. Simulations show that
extrapolation errors can be much improved after only 1 iteration. Further iterations
do not improve the extrapolation error a great deal. Therefore, from the practical
point of view, 1 iteration is sufficient to obtain better result.
From the reconstructed image shown in Figs. 4.9(c) and (d), u may lead to a
conclusion that iterative algorithm is not very helpful in the application of radar
imaging or extrapolation of scattered fields into unavailable bands. It has to be
pointed out that the iterative method is necessary only when the data sequence can
not be modeled properly from the single band.
The application of the linear model or all-pole model to the extrapolation of
scattered field into unavailable bands is based on the assumption that the scattering
properties of the scattering mechanisms are not a function of frequency. This assump­
tion is valid only when the wavelength is much smaller than the dimension of the tar­
get. The test target used is a 1:100 scale model B-52 airplane, and the frequency
range used is from 6.0 to 17.5 GHz. In a real imaging system, however, the fre­
quency coverage will not be so broad. For example, the center frequency and
bandwidth of a practical radar imaging system can be about 2 GHz and 200 MHz
respectively. The scattering mechanisms and their scattering properties will not
change a great deal over that frequency coverage. Therefore, the extrapolation algo­
rithm with or without iteration will probably work better in practice than in this test­
ing system.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 104-
The previous examples show the extrapolation of the scattered field into the
exterior bands for a given transmitter and receiver. One may ask if the same algo­
rithm can be used to extrapolate the complex field over a given angular aperture into
the exterior aperture. If the object consists of discrete point sources and each point
source radiates isotropically, then the source model can be considered as an all-pole
model. One may deduce the field outside the aperture from the known field in the
given aperture. However, the all-pole model is not valid for a metallic object over
all space, because the locations of the scattering centers and their scattering strength
are highly dependent on the aspect angle. The shielding effect of the conductor
prevents the information of the object in the shadow region from being detected.
Therefore, the application of the linear prediction method for extrapolation of the
scattered field in a given aperture to a wide exterior aperture is not recommended.
The image quality of the reconstructed images using extrapolation data looks as
good as that obtained from the whole band data. However, there is a difficulty in the
definition of "good". The quality of an image is subjective. To measure the effect of
the extrapolation on the reconstructed image, one may subtract the image obtained
from extrapolation data from that of the whole band data, pixel by pixel, using com­
puter and then compare the difference. However, this kind of measure is not effective
because the significant features of an image can be extracted only from the whole
picture but not from a single pixel or a single line of the image. Even though the
difference obtained by subtracting pixel by pixel may not be small, the images may
still look similar. On the contrary, the extrapolation errors in the frequency domain
can be defined clearly because they can be compared with the experimental data
exactly. The development of a technique which quantitatively measures the extrapola­
tion algorithm in object space would significantly enhance the analysis of this
research.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 105-
CHAPTER5
RADAR CROSS SECTION MANAGEMENT STUDIES
EMPLOYING MICROWAVE DIVERSITY IMAGING
A s mentioned in Chap. 2, radar cross section (RCS) is a measure of the equivalent
size of a target as seen by the radar. RCS is a function of aspect angle, frequency,
and polarization. For certain applications, it is desirable to enhance the RCS over
some specific range of aspect angles. For example, strong return signals are required
to track a missile or an aircraft in flight. In contrast, for other applications it is often
desirable to reduce or minimize the RCS over specified spectral and angular windows
so the target is less likely to be detected.
RCS management is a general term for obtaining the RCS of a scattering object
by manipulating the distribution and strengths of the hot spots or flare spots of a tar­
get over prescribed spectral and angular windows and states of polarization. Hot
spots represent all those areas that have major contributions to the received scattered
fields. There are several ways to determine the site of hot spots. An experienced
engineer can roughly point out the possible locations of the hot spots; a computer
code can be used to estimate the locations of the scattering centers and their scatter­
ing strength if the equations that describe the surface of the object can be formulated.
However, all such estimates need experimental verification.
Microwave diversity imaging systems using angular, wavelength, and polariza­
tion diversity have been discussed in Chap. 3 and have been used to obtain images of
metallic objects with nearly optical resolution. It was stated in Chap. 3 that the pixel
intensity or brightness of the reconstructed image is proportional to the mean scatter­
ing strength at the corresponding position averaged over the specified angular win­
dow and spectral window. This fact suggests microwave diversity imaging can be
employed to determine the hot spots and to study RCS management.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 106 -
Two steps are usually involved in the study of RCS management. First the hot
spots must be identified. Certain techniques are then utilized to manipulate the hot
spots. Fine resolution of object detail in the reconstructed image is essential for
locating the hot spots of an object over the prescribed spectral and angular windows
and the given states of polarization of the transmitter and receiver. The range resolu­
tion of a frequency diversity imaging system is proportional to the bandwidth of the
measurement system. If the bandwidth is too narrow to give acceptable resolution,
the algorithm for the extrapolation of data available in multiple restricted bands or in
a single narrow band developed in the previous chapter can be applied to locate the
hot spots.
Previously, radar workers usually treated the RCS from the detection perspec­
tive. Therefore, techniques of reducing RCS were developed to reduce the signal to
noise ratio (S/N) so that the target can more easily elude detection from radar. How­
ever, as will be shown later a low S/N is not a sufficient criterion for eluding detec­
tion when microwave diversity imaging systems are employed where signals from
many broad-band sensors are combined to form an image. In this dissertation we
will consider therefore the RCS management from the imaging point of view, that is,
we consider principles and techniques for making the image unrecognizable by radar.
In this chapter we will introduce a new term, "diaphanization", and describe how to
use microwave diversity imaging to diaphanize an object. Effects of Gaussian noise
on the reconstructed image will be discussed. Some fundamental concepts for distort­
ing the image will be proposed.
5.1 Diaphanization
According to the Webster dictionary, “diaphanous” means:
1.
so fine or gauzy in texture as to be transparent or translucent
2.
vague or indistinctive, airy.
Accordingly, we introduce a new term “diaphanization”, which can be used to
describe:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 107-
1.
All techniques of reducing RCS in order to render a target invisible to radar
detection.
2.
All techniques of obscuring the image so that the image is vague, indistinctive
or not recognizable by an imaging radar
Therefore, diaphanization is a technique which will influence either detection jor
image formation and recognition.
From the detection perspective, diaphanization will generally lower the target’s
signal to noise ratio (S/N), thus decreasing the likelihood of detection [76]. There­
fore, it is desirable to minimize the RCS. From the imaging point of view, however,
low S/N is not a fatal problem for microwave diversity imaging systems. It will be
shown later that microwave diversity imaging is very robust to Gaussian noise, even
though the noise is stronger than the signal. Therefore, techniques that distort the
reconstructed image to make it unrecognizable should be considered. In the follow­
ing, we will demonstrate how to apply microwave diversity imaging to diaphanize an
object and to explain the motivation for studying image distortion.
The procedure of utilizing microwave diversity imaging to diaphanize a target
for prescribed spectral and angular windows is as follows:
1.
Obtain the image of the target before diaphanization.
2.
Identify the hot spots which are the bright portions in the image.
3.
Use certain techniques to diaphanize the object, i.e. to minimize reflection from
the hot spots.
4.
Obtain the image of the target following diaphanization.
5.
Compare the images before and after diaphanization.
An example will be given below to demonstrate the above procedure. In the
following experimental example, 201 equal frequency steps covering the 6.0 to 17.5
GHz range were used to obtain the frequency response of the object for different
aspect angles. The magnitude of the scattered fields are measured relative to a refer­
ence cylinder. The RCS defined in this chapter is the square of the measured relative
field and therefore the unit of RCS is a relative level but not an absolute value. The
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 108-
measurement is referred to as co-polarized when both the transmitting and receiving
antennas have the same sense of circular polarization; and is referred to as crosspolarized when the transmitting antenna and receiving antenna have an opposite sense
of circular polarization.
The test object, a metalized 1:100 scale model of a B-52 airplane with 79 cm
wing span and 68 cm long fuselage was mounted on a computer-controlled positioner
situated in an anechoic chamber environment. A sketch of the B-52 and the recon­
structed image using data collected in an angular window of 90° extending from
head-on to broadside in 128 looks with cross-polarized waves are shown in Figs.
5.1(a) and (b) respectively. The intensity or brightness of each pixel is proportional to
the average scattering strength of that portion over the angular and spectral windows.
The bright spots are the hot spots of the target. To reduce the RCS, we cover
the hot spots with broad-band absorbers (Emerson and Cumming AN-72). Portions of
the B-52 model covered with broad-band absorber are indicated by the dark lines in
the sketch of Fig. 5.1(c). These parts consists of the right hand sides of the fuselage,
all edges of the right wing, and the right sides of the two right engines and fuel tank.
The RCS of an object is usually reduced after diaphanization, but it is not reduced
for every frequency because the scattered field is a result of interference and is a
function of frequency. Therefore, it is not proper to measure the effectiveness of a
diaphanization on RCS at a specific frequency if a prescribed spectral range is of
interest. We have adopted a more reasonable measure — that of comparing the
mean scattered fields before and after diaphanization. The mean RCS is obtained by
averaging the RCS at each frequency over the prescribed spectral window for a given
aspect angle. The mean RCS versus aspect angle before and after diaphanization
using cross-polarized waves are shown in Figs. 5.1(d). From these plots, we find that
the RCS has been reduced for most aspect angles. The reconstructed image after
diaphanization are shown in Figs. 5.1(e). The images before and after diaphanization
have the same dynamic display range. One can see that diaphanization reduces the
brightness of the hot spots. However, if we lower the dynamic display range of the
post diaphanization image, the image becomes as shown in Figs. 5.1(f). Clearly, the
nature of the target is still recognizable. From a detection point of view, covering
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 109 -
the hot spot portions did reduce the RCS. However, from the imaging perspective,
the image is still recognizable.
5.2 RCS Reduction
Reduction of a target’s RCS will lower the S/N and therefore decrease the
detection probability and increase the false alarm rate [76]. For example, Shown in
Fig. 5.2 is the probability of detection for a sine wave in noise as a function of the
S/N and the probability of false alarm. Furthermore, the detectable distance of a tar­
get will be halved if the RCS of the target is reduced by 12 dB, as derived from the
radar equation [76]. Therefore, it is desirable to make the RCS as small as possible.
The practical question persists, to what extent can the RCS be reduced? The follow­
ing RCS reduction numerical examples are instructive in the answering of this ques­
tion.
At high frequency, the scattered field at a given aspect can be attributed to
several discrete sources or scatterers as stated in Chap. 2. The RCS at a given fre­
quency is then
a ( k ) = I ' Z ' f c e j2kri I'2 = 1
i
1
2
( 6. 1)
i
where a,-, r ; , and <{>,■ are the RCS, relative range, and relative phase of the i -th
scatterer respectively. It is noted that Eq. (6.1) is a coherent summation. In general,
a, and rv are aspect dependent but are independent of frequency or only vary slightly
with frequency as discussed in Chap. 2. However, the total RCS fluctuates rapidly
with respect to aspect angle and frequency because of the coherent sum of the phase
term ejt>i, which is highly frequency and aspect dependent. If the RCS of interest is
analyzed over a wide spectral window, it is not proper to compare the RCS before
and after diaphanization at a single frequency. It is far better to compare the mean
RCS averaged over the window of interest. The phase angles are then equally prob­
able because of the frequency stepping or sweeping and the mean RCS can be statist­
ically characterized as the noncoherent sum
<*m = Z
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
<6-2)
- 110-
45
before
amplitude
in dB
35
after
25
15
5
128
i i ne number
d
Fig. 5.1 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- Ill -
®
f
Fig. 5.1 (a). Sketch of the B-52 and the angular window.
(b). Reconstructed image of B-52
(c). Sketch of the portion (dark area) covered with broad-band absorber.
(d). The mean scattered field versus aspect angle before and after diaphanization.
(e). The reconstructed image after diaphanization displayed with the
same dynamic range as that of (b).
(0- Reconstructed image display with a lower dynamic range.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 112-
The mean RCS pattern then represents the averaged scattering characteristics of the
target as a function of aspect angle averaged over the specified spectral window.
The noncoherent sum is helpful in illustrating the extent to which the mean RCS
can be reduced. Numerical examples have been given in [33] to show the RCS
reduction number game. Assume a simple target consisting of three contributors, a
strong one with 200 m 2 in area and the remaining smaller two 20 m 2 in size.
Several strategies were considered and the resultant RCS reductions were compared.
The effect of reducing the returns of one or all of the contributors, the effect of
working harder on the dominant scatterer, and the effect of selective elimination of
scatterers are shown in Table 5.1 to 5.3 respectively. If all three contributors have
the same amplitude, the effect of reduction is as shown in Table 5.4.
The examples above illustrate the increased difficulty associated with reducing
the RCS of an object when multiple contributors are present. The total RCS reduc­
tion can have the expected results only if all the contributors have been reduced
effectively. For this reason each decrement in RCS is obtained at a successively
higher cost.
Techniques of reducing the RCS can be grouped into three classes [21]: cover­
ing the object with absorbing material or substituting metallic parts with resistive or
composite material; shaping the target; and impedance loading. The first method is
based on absorption of the incident radiation. The second method is based on
redirecting the incident waves, i.e., scattering the incident illumination into as wide a
solid angle as possible or into a direction away from the probing system. The third
method, also known as passive cancellation, is based upon the concept to introduce
an echo source whose amplitude and phase can be adjusted to cancel another echo
source [33]. In this paper, we will give examples showing how microwave diversity
imaging is used in RCS management and diaphanization studies of a complex
object.
It is important to emphasize that RCS reduction is a study of compromise in
which advantages are balanced against disadvantages [33]. For example, a reduction
in RCS at one aspect is usually accompanied by an enchancement at another aspect if
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 113 Tablef-!
Effect of a Dominant Scatterer [33]
R educe
R educe
0t by
10 dB
0 |. 02
by lOdB
20
20
20
20
2
20
20
2
2
U ntreated
200
0i
02
Reduce
<72 , 0j
by 10 dB
0 \,
O}
20
TOTAL
240
60
42
24
dB reduction
0.0
6.0
7.6
10.0
Tablet-2
rooi
Effect of Working Harder on the Dominant Scatterer L J
R educe a t . 02 R educe ot by 15 dB.
02 by 10 dB.
by 13 dB .
ot by 10 dB
02 by 10 dB
U ntreated
R educe
at by
15 dB
02
200
20
20
6.3
20 .0
20.0
6.3
2.0
20.0
6.3
2.0
_10
T O TA L
240
46.3
28.3
10.3
0
7.1
9.3
13.7
0\
02
dB reduction
Table f '3
Effects of Selective Elimination ofScatterers ^ J
Untreated
Ot
Elim inate
o\ and ot
0\
200
20
20
0
20
_20_
200
0
_0
T O TA L
240
40
200
dB reduction
0.0
7.8
0.8
E lim inate
o\
O'.
Table 5-4
Effect of Reduction When AH Have the Same A mplitude £33]
R educe
ot
by
R educe
Reduce
Ot, 02
01, 02. Ot
by 10 dB
Untreated
10 dB
by 10 dB
20
20
20
2
2
20
TOTAL
1“
60
2
20
20
42
24
6
dB reduction
0.0
1.6
4.0
10.0
Ot
O'.
Ot
•
2
2
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 114 -
the surface of the object is reshaped. The use of absorbing material will increase the
weight, volume, and cost of manufacture and maintenance. Therefore, trade offs
have to be made according to the requirements for the system design and its mis­
sions.
5.3 Effect of Gaussian Noise on Microwave Diversity Imaging
As shown in Fig. 5.2, a low S/N will decrease the detection probability and
increase the false alarm rate. For a practical radar detection system, S/N above 15 db
is usually required. It is also known that coherent summation will produce a coherent
gain or superposition gain. That is, if a signal contaminated by Gaussian noise are
coherently summed n times, the resultant S/N will increase n times [76]. The recon­
struction algorithm of microwave diversity imaging involves obtaining the filtered
range profiles and back-projecting them in the image plane. Both procedures involve
coherent summations and hence result in coherent gain. Therefore, the robustness of
microwave diversity imaging to Gaussian noise is expected.
Microwave diversity imaging usually uses stepped frequency response measure­
ment For each frequency step, the bandwidth of the receiver is kept the same. The
standard deviation of the Gaussian noise at each frequency step is therefore the same.
Under the high frequency approximation, at a specific aspect the field scattered
from a metallic object can be expressed as the superposition of the scattered fields of
N discrete scattering centers such that the scalar fields at frequency f m is given by
Es i f m) = E ai exP0 2jt/m ri)
i=1
2
(6-3)
where a,- is the scattering strength of the / -th scattering center, r-t is the differential
range of the i -th scattering center with respect to the reference point, / j and / 2 are
the lowest and highest frequencies of the measurement system respectively, Np is the
number of frequency points. Assume the noise power level of the receiver at each
frequency step is N, then the S/N due to the i -th point scatter is Ia,12 / N . Assume
the point scatterers are well separated apart so that at range rt the contributions due
to the side-lobe of other scatterers can be neglected. The signal (sampled value at r,-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 115-
Q 2 9 0 0 0 0 0
0.9999
0.9995
0.999
0.998
0.995
0.99
0.98
|
0.95
U
I
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
Probability ot
to n * alarm —
10“ 1or1 lO^Kr'iaW'
0.05
4
6
8
10
12
14
(V * ),, jignal-to-noist ratio. 48
16
18
20
Fig. 5.2 Probability of detection for a sine wave in noise as a function of
the S/S and the probability of false alarm.
in the FT domain or range profile) to noise ratio will become Np Ia,-12 /N since the
Fourier transform is a process of coherent summation. Therefore Np is the coherent
gain or superposition gain for this example.
Back-projection is a process of coherent summation of the contributions of the
back-projection lines obtained from the filtered range profile of each aspect angle.
This additional coherent process will make the reconstructed image more robust to
the Gaussian noise. If the object consists of isotropic point scatterers, i.e., the scat­
tered fields of these point scatterers are isotropically distributed, then the overall
coherent gain will be A!aNp if data from Na aspect angles are used to reconstruct the
image. In practice for a metallic object for example, the locations and scattering
strengths of the scattering centers and end points of edges will change or migrate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 116 -
when the object changes aspect (rotation) during data acquisition. Therefore, the sig­
nal (image intensity at a pixel) to noise ratio in the image space cannot be expressed
in the above simple form.
The following example illustrates the effect of Gaussian noise on the recon­
structed image.
Figure. 5.3(a) shows the mean RCS after diaphanization of the B-52 scale model
versus aspect angle. We used the computer to generate Gaussian noise with two lev­
els N i and N 2 (see Fig. 5.3(a)) and add those to the measured fields to get two sets
of noisy data. The corresponding S/N levels of N i and N 2 316 about 3 dB and -2 dB
respectively relative to the mean RCS. The frequency response of the first aspect
angle (broadside view) without and with noise levels N j and N 2 are shown in Figs.
5.3(b), (c), and (d) respectively. The range profiles corresponding to the data in Figs.
5.3(b), (c), and (d) are shown in Figs. 5.3(e), (f)» and (g) respectively. If we stack
the range profiles of the first 64 view angles together in a perspective representation,
the range profiles without and with the two levels of Gaussian noise will appear as
shown in Figs. 5.3(h), (i), and (j). The images reconstructed from these three sets of
range profiles are shown in Figs. 5.3(k), (m) and (n) respectively. The results
presented in Fig. 5.3 show clearly that Gaussian noise affects the frequency response
most, the range profile less, and the reconstructed image least because of coherent or
superposition gain.
5.4 Fundamental Concepts for Distorting an Image
From the above examples, it can be seen that low S/N or strong Gaussian noise
is not a sufficient criterion for totally obscuring the images. In order to develop tech­
niques for making the image unintelligible or obscured, one must resort to under­
standing of the scattering mechanism and the reconstruction algorithm involved.
Radiation can originate from several locations on a scatterer. These include
excitation regions, impedance loads, sharp bends, and open ends [38]. The finite
passband data and finite range resolution make the range profile accentuate the
discontinuities in reflectivity function, and the image can be interpreted as a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 117 -
20
amplitude
in dB
no i s e l ev e l N
15
10
n o i s e l ev e l N,
5
0
1
32
64
96
128
1ine number
a
Fig. 5.3 (caption see page 121)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.0
8.9
11.2
14.6
17.5
FREQUENCY IN GHz
20
LU
O
Z>
h-
6.0
8.9
11.2
14.6
17.5
FREQUENCY IN GHz
c
I
8.9
11.2
14.6
17.5
FREQUENCY IN GHz
Fig. 5.3 (caption see page 121)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 119 -
MAGNITUDE
4.0
3.0
1.0
-130
-6 5
0
65
130
DISTANCE IN cm
MAGNITUDE
4.0
—
1-----
‘i
0
65
3.0
2.0
1.0
-130
4.0
MAGNITUDE
1—
-65
130
DISTANCE IN cm
" ■■■ r - - —r — .... . r
3.0
2.0
1.0
0
-130
-65
0
65
130
DISTANCE IN cm
s
Fig. 5.3 (caption see page 121)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 120 -
2 . 2e*0l
•€.38e+0i -3 .3 2 e» C i‘
8.0:j'
3 .3 * * 0 t
. 35e*vl' " 3 . 32e*0l*
C. 0*+00
j
Fig. 5.3 (caption see page 121)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fig. 5.3 (a). Averaged scattered fields of the scale model B-52 after diaphani­
zation versus aspect angle and the generated Gaussian noises with
level iVj and N 2.
Frequency response of the first view angle (broadside)
(b). without adding noise, (c). with noise level N h and (d) with noise level N 2.
(e). The range profile of (b); (f). The range profile of (c);
(g). the range profile of (d);
The stacked range profiles of the first 64 view angles (h) without adding
noise; (i) with noise level N \; and (j) with noise level N 2.
The reconstructed image (k) without adding noise; (m) with noise level
and (n) with noise level N 2.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 122 -
collection o f the discontinuities of reflectivity function [5]. The reconstruction algo­
rithm followed by include steps: obtaining filtered range profiles followed by backprojection. Both steps as pointed out earlier are coherent summation processes.
Based on these observations we will propose some fundamental concepts or rules for
obscuring the image.
1.
Creation of artificial discontinuities: The image is a collection of the discon­
tinuities of the reflectivity function. For a set of given objects, their images can
be obtained and stored to act as the data base for subsequent recognition pur­
pose. If unexpected discontinuities were created, the reconstructed images of the
target will be different from their original ones. In that case, visual recognition
will be more difficult or may fail.
2.
Creation of multiple reflections: Locations of peaks in the range profile due to
multiple reflections correspond to the differential ray path between the multiple
scattering points on the object. Such peaks correspond to as physical detail on
the object and help therefore to distort and obscure information about object. If
the object is shaped so that multiple reflections are deliberately produced to be
the dominant contributions to the scattered fields, then the image might be
obscured or distorted.
3.
Manipulation of local reflectivity: Here we can make the reflectivity at a given
point a function of aspect angle and frequency or randomly modulate it in time.
The contributions to a pixel from different filtered range profiles obtained at
different aspect angles are added coherently. If the reflectivities of a point
obtained at different aspect angles have the same phase, that image point will be
intensified after back-projection. If the phase of reflectivity at that point is ran­
domly varied for each frequency and aspect angle, the range profile might pro­
duce random peaks and the peaks in the range profile corresponding to those
reflection points might become smaller because their reflectivities at each fre­
quency is not in phase. After back-projection, the intensity at that image point
might become smaller and the image might be contaminated by random noise.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 123 -
In the following chapters we will apply the above ideas to study some diaphani­
zation techniques which might be useful in reducing the RCS or distorting the images
and making them intelligible.
5.5 Discussions
There are three hierarchies in high resolution imaging radar. By frequency
diversity, one can obtain the range profile, which gives the range information of the
hot spots at that aspect. However, there are ambiguities in determining the hot spot
positions in the directions of the cross-range and the rotational axis. By applying
angular diversity in the azimuthal direction, the positions of the hot spots in the
direction of cross-range can be located. However, the images obtained are projective
images. The locations of the hot spots in the direction of the rotational axis are still
ambiguous. If another angular diversity in the elevation direction is employed, by
applying the weighted projection theorems [9], a 3-D tomographic image can be
obtained and the hot spots can be pinpointed more accurately. Frequency swept
tomographic imaging of 3-D perfectly conducting objects has been demonstrated by
computer simulation [9] and experiment [60]. It was stated that by multiplying the
3-D space data by a weighting function, a tomographic image at any selective plane
can be obtained. The analysis is based on the the PO approximation and is valid for
point-scatterer objects. However, a more convincing interpretations for 3-D imaging
of a complex shaped metallic object needs to be developed.
The RCS of a target at a certain frequency can be very small due to the
coherent interference of the scattering centers. In that situation, the detectability of
the target will be very small. But its RCS may not be small at many other frequen­
cies. The detection probability and false alarm shown in Fig. 5.2 was analyzed based
on a sinusoidal wave at a single frequency. If frequency diversity is employed, The
modification methods for detection criterion utilizing the frequency diversity proper­
ties so that the radar is more powerful in detection are worth investigating.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 124 -
CHAPTER 6
DIAPHANIZATION BY ABSORBER COVERING
AND TARGET SHAPING
A s defined in the previous chapter, diaphanization refers to techniques of reducing
RCS and techniques of obscuring an image. In this chapter, we will consider and dis­
cuss RCS modification by covering the target with absorbing iayers and by target
shaping.
Metallic bodies are good reflectors and scatterers of electromagnetic waves.
However, high reflectivity is not a desirable property in some applications. For exam­
ple, the large metallic structure on ships or airplanes makes them easily detected by
radar, masts, beams, towers, hangers, and other shipboard structures are not only obs­
tacles to the functioning of communication systems, but can also be sources of
interference, which may degrade the performance of shipboard communication gear
and distort antenna radiation patterns.
Application of radar absorbing material
(RAM) to metallic bodies can reduce the RCS and eliminate the undesired interfer­
ence by partially absorbing incident EM energy, suppressing thereby reflections and
scattering.
To reduce the RCS of a target with a coating of RAM is, in effect, to reduce the
reflection coefficient of the air-RAM interface. Single-layered and multilayered
coated surfaces have been analyzed using transmission line modeling [21], and will
be summarized in Sec. 6.1. But this approach fails to predict the correct field over a
wide range of aspect angles, since the transmission line method does not consider
diffraction phenomena. Scattering from resistive sheets and resistive strips has also
been reported in the literature [76], a high frequency solution for diffraction by a
strip with two arbitrary surface impedances when illuminated at edge-on incidence is
obtained by a spectral extension of the geometrical theory of diffraction [78]. A solu­
tion for the problem of EM scattering by a half pilane with two different surface
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 125 -
impedances for both normal and oblique incidence was developed [79]; an asymptotic
high frequency estimation of monostatic RCS of a finite planar metallic structure
coated with lossy dielectric was made theoretically and compared with experimental
results [80]. In this chapter, we do not attempt to numerically calculate scattering
from a plate covered with RAM using the above developed formulations. The electri­
cal properties ( e, ( i ) of the RAM used differ from the theoretical assumptions of the
cited literatures. Therefore, we will concentrate more on experimental studies.
RAMs can be divided into four groups: magnetic material absorbers, resonant
absorbers, broad-band absorbers, and surface current absorbers [81]. In Sec. 6.2, we
will briefly discuss the properties of these four RAMs. Experimental results of the
RCS and images of metallic objects before and after covering by RAM will be
presented and discussed in Sec. 6.3.
Target shaping is another mean for reducing RCS. The basic notion of target
shaping is different from covering by absorbing material in that it is nonabsorptive.
A commonly observed phenomenon is any attempt to reduce RCS by a nonabsorptive
method over a range of aspect angles is promptly and assuredly met by an enhance­
ment of the RCS over some other range of aspect angles. The above phenomenon is
sometimes considered as the "radar cross section conservation" principle, although it
is not a principle in the strict sense of the word [82]. In some applications, only a
certain range of aspect angles is of interest. Therefore, the object is designed to
minimize the RCS within the angular window of interest.
To investigate the effect of target shaping on the RCS, it is helpful to compare
the RCS of simple objects over prescribed angular and spectral windows. The RCS
of some simple objects have been analyzed in [21]. Radar cross section reduction
using cylindrical segment has been reported [82]. In Sec. 6.4, we will compare the
RCS for objects of various shapes and state the basic rules for target shaping con­
siderations.
We have concluded in Chap. 3 that the reconstructed image represents "discon­
tinuities in the reflectivity function and its derivatives." If we coat the target surface
inhomogeneously, i.e, we coat the surface only in certain areas, then we creat
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 126-
artificial discontinuities in the reflectivity function. These discontinuities can be
reflected in the reconstructed image. If the boundaries of the covered regions and
conductors are suitably shaped, the total scattered field may be more effectively
reduced. In Sec. 6.5, we will examine how the combination of absorbing material
covering and shaping affects RCS and imaging.
6.1 Absorber-Covered Bodies - Theoretical Considerations
Consider a metallic object covered with a homogeneous absorbing material with
complex permittivity E2, permeability (J.2, and illuminated by a source 7q as shown in
Fig. 6.1. Vm and Sm denote the volume and the boundary surface of the conductor,
Va and Sa denote the volume and the outer surface of the absorber respectively, and
V0 denotes the volume of the flree space. The general formulation of Maxwell equa­
tions for this structure lead to
V0
To
f
Fig. 6.1 Geometry of an absorber covered body illuminated by a plane wave.
( V2 + Icq )
(r*) = T0(f)
( V2 + k l ) & z ( f) = 0
where k§ = (D^Eo, k \ = w2M-2e2>^1
space,
for 7* in free space
(6.1)
for r e Va
(6.2)
is the total electric field at point r* in free
is the total field at point r* inside the absorbing material. The solution
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 127 -
must satisfy the following boundary conditions
nm( t ) x ^ 2(f) ~ 0
for r*on Sm
na(r)i<h? 2 ~
(n) = 0 t on SA
na(r)x(!Pl ( t ) - i t t2(t)) = 0 to n SA
na ( t ) i \ h A ~ H i# i (/*)) = 0 t o n SA
na<?)x(A(?) -
(r*)) = 0 t o n SA
where nm (t) is the unit vector normal to the surface Sm at t , and ha (t) is the unit
vector normal to the surface Sa at t .
In the absence of the metallic object and
absorbing material, that is, the whole space being free space, the solution of
Maxwell’s equations is the incident wave and is denoted by E?(t). The scattered
field from the absorber covered body is then defined as
2*0*) = ^ < ? )-& (? )
(6.3)
In general, the exact solutions of
(t) are not available except for those
boundary surfaces which coincide with the surfaces of curvilinear coordinates. Exact
solutions of the scattered field from a coated sphere and coated infinite cylinder have
been analyzed in [21]. In the following, we will review the properties of a semi­
infinite plate covered with RAM and then use the PO approximation to study the
absorber covered body.
6.1.1 Semi-infinite Plate
The scattering problem of a plane wave incident at an arbitrary angle on a
medium which consists of an arbitrary number of plane layers, each of which is
homogeneous and has specific dielectric properties, has been analyzed in [21]. A sin­
gle homogeneous layer which covers an infinite conducting plate in otherwise free
space is a special case and its geometry is shown in Fig. 6.2. In this figure, Gj is the
angle of incidence and reflection,
and (J.2 are the complex permittivity and com­
plex permeability of the layer respectively, and d is the thickness of the layer. By
using the plane-wave and transmission line analogy or by applying suitable boundary
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 128 -
conditions, the reflection coefficients Rp and Rv can be shown to be [21]
a =
Fig. 6.2 Geometry of a semi-infinite conducting plate covered with a
homogeneous plane layer.
R _ K
_ ~Hy
-T1o-yz!tanK2d
H‘
T|0-yz§ tamc2d
p
/\y
_
E; _ -r\Q-jzv2vmK2d
™^
EJ
where tj0 = ViVeo ^
(6. 5)
710-yz$tanK2d
characteristic impedance of free space, k2 =
*s the
complex wave number of the layer, the subscripts p and v represent parallel and per­
pendicular polarizations respectively, z2 and z£ are the equivalent complex charac­
teristic impedance of the layer and are defined by
z v2 = —
a / ^ " =
^ l/e2
cos02 \ e2
V 1 - (MoEo/^e^sin^!
(6.6)
for an incident electric field vector perpendicular to the plane of incidence, and
z \ = cose2V
f
= V lV ^V l - ((ioeo/M-2e2)sin20i
(6.7)
for an incident electric field vector lying in the plane of incidence. 02 in Eqs. (6.6)
and (6.7) is the refraction angle in the layer (medium 2 in Fig. 6.2).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 129-
For a lossy material, the permeability and permittivity are complex and are
given by
M-2= 1*2 + iV-2
(6.8)
e2 = e2' + ye2"
(6.9)
If the operating frequency / , incident angle 8^ permittivity e* permeability
and
thickness d of the absorbing layer are given, the reflection coefficient can be deter­
mined. It is noted that the reflection coefficients for vertical polarization and horizon­
tal polarization are generally different at oblique incidence. Therefore, the reflected
wave of a circularly polarized incident wave is, in general, ellipdcally polarized.
Shown in Fig. 6.3 are examples of the magnitude and phase of R v and R p
versus frequency for a given set of parameters with
= 1 + y'0.4, p.2 = 1 + y'O,
9t = 0° (normal incidence), and thickness d=0.8 cm. R v is equal to R p at each fre­
quency for this special case of 9i=0°. For the other incidence angles, IR p I is in
general not equal to l/?wI if the layer is lossy. Shown in Figs. 6.4(a) and (b) are the
magnitude and phase of R v and R p versus the incidence angle with parameters
e2 = 1 +j 1 .0 ,112 = 1 + yO, /= 1 1 GHz, and d - 0.6 cm. The parameters chosen are to
simulate the broadband absorber to be used later. It is seen that the magnitude and
phase of l/?v I are not equal to those of \RP I for every incident angle. The magni­
tude and phase of the reflection coefficients for the co-polarized waves and crosspolarized waves with the same parameters are shown in Figs. 6.4(c) and (d) respec­
tively.
6.1.2 Physical Optics Approach for an Absorber-Covered Metallic Object
The reflection coefficient derived in the previous sub-section is valid only when
the plate is infinite and when the receiver is in the forward direction, i.e., the normal
of the plate is in the direction bisecting the angle between the transmitter and
receiver. The problem of an arbitrarily shaped metallic object covered with an
absorber is very difficult to analyze. However, if the radii of curvature of the object
is large compared to a wavelength, the physical optics approximation may be used to
analyze the problem.
m ---------------Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 130-
logl/?v l =logl/?*
0.75
log scale
U
TJ
0.50
c.
I -30:
■ 0.25
linear scale
J O .O
-40
6.0
8.9
11.8
14.7
17.5
frequency in GHz
(a)
180
eo
<u
TS
e
93
t/j
2
"a.
-90
-180
6.0
8.9
14.7
11.8
frequency in GHz
17.5
(b)
Fig. 6.3 Reflection coefficients Rp and Rv of a semi-infinite
plate covered with an absorber layer versus frequency
The parameters used are e2 = 1 + y 1, ji2 = 1 + y'0, 0 1 = 0° (normal
incidence), and 4=0.6cm.
(a). Magnitude of Rv and Rp in linear scale and in log scale.
(b). Phase of Rv and Rp .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0
•88
-44
0
44
incident angle in degree
(a)
88
180
SO
lR
-180
0
-88
44-
88
incident angle in degree
(b)
0.75
■38
.B
0.50
K
,
IRa l
-88
0
88
44
incident angle in degree
180
i
-90 ■
-180 L .
-88
,
-44
0 . 4 4
incident angle in degree
—
88
Id)
Fig. 6.4 M agnitude and phase o f
Rv
and
Rp
versus incident angle.
(a). M agn itu d e o f R v and R n .
(b). Phase o f
Rv
and
Rp .
(c). M agn itu d e o f reflection c o e ffic ien t for c o -p o la r ize d and c r o ss-p o la rized w a v es.
(d ). P h ase o f reflectio n c o e ffic ien t for c o -p o la r iz e d and c r o ss-p o la rized w a v es.
T h e param eters u sed are: e : = 1 -i- ; 1 . |4; = 1 + ; 0 , / = U
G H z.
and d=0.6cm.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 132 When an absorber-covered body is illuminated by a plane wave, a field will be
induced inside the absorber. If only the field external to the absorber is of interest,
the original structure can be replaced by an equivalent structure through the
equivalence principle [45]. The two equivalent structures are shown in Fig. 6.5,
where the internal field has been set to zero, and an equivalent electric surface
current density Js and an equivalent magnetic surface current density Ms have been
applied to the surface SA. To support the original field outside the absorber, Le., to
make these two structures equivalent, die equivalent current density must satisfy the
relations
(6.10)
Ts = n x l F
and
(6.11)
where n is the outward normal, and I F and FF are the total original fields over S q.
The equivalence principle does not help us in solving the problem, because
eF
and FF remain unknown. It only helps us to view the problem from another per­
spective. However, if IF and FF can be obtained or approximated through other
reasonable assumptions, then the equivalence principle will be a very valuable con­
cept.
If we assume that the reflection on the surface takes place as if the point of
reflection lies on a fictitious infinite tangent plane, then the total field at a point r* on
the boundary surface SA could be expressed as
£^(7=*) = £*> *)+ £*(?*)
= t T* F ) + ? Ty<F) + '£rv:<F) + £ rTM<F)
(6.12)
i F ( T ) = t f 'c n + f f r <F)
= /T * ^ ) +
+ #'*(/'•') + FT™ {t*)
(6.13)
where the superscripts i and r represent the incident and reflected waves respec­
tively, and the superscripts TE and TM represent the TE and TM components
respectively. The incident field and the reflected field on the boundary surface can
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 133-
F-
Eq
'•
? = ri x f f ‘
S*
l? ,= ? x n
Fig. 6.5 The two equivalent structures of an absorber covered body illuminated by a
plane wave.
be related to the reflection coefficients Rp and Rv and are defined as
F 7* = R vg iTE
(6.14)
s 'x F ™ =Rp s ix ? TU
(6.15)
where s l and s T are the unit propagation vectors of the incident wave and reflected
wave respectively. Rv and Rp are obtained as if the surface at the reflection point
was an infinite tangential plane.
Their expressions are shown in Eqs. (6.4) and
(6.5). It is noted that Rp and Rv are functions of frequency, angle of incidence,
absorber’s composition and the object’s geometry. If the surface of the absorbercovered body and its equivalent surface impedance can be mathematically described,
the reflection coefficients Rp and Rv and accordingly the total fields ET, H T and the
equivalent current densities Ts ,
can then be determined. The back-scattered fields
can finally be calculated by Eqs. (2.23) and (2.24).
Without loss in generality, consider an incident field propagating in the z direc­
tion and describe the surface of the absorber-covered body by the equation
z' = f ( x ' , y ' ) . The geometry is the same as that shown in Fig. 2.3 with the notable
difference of the absorber covered surface substituted for conducting surface.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 134-
Expressions for the surface’s normal vector n and the surface element area ds' are
shown in Eqs. (2.36) and (2.37) respectively. The expressions of the back-scattered
field for various incident field polarizations can be found in [21]. For the special
case of an x -polarized incident wave, that is,
the back-
scattered field is given by [21]
.
(d fld x 'j1R. - 0 f / d y ' ) %
,
0 f / t e ) 2 + 0 // 3 y ') 2
(Rp+Rv)(dfldx'Xdf/dyf ) ,
■ej2kof dx'dy'
@ f/dx')2 + (df/d y')2
(6.16)
-Wo
=■ £= = [0^+ 0^]
V4jt/?o
(6.17)
Expressions of Rp and Rv are functions of the
incidenceangle6;, which
expressed in terms of the integration variables x' and y' asgivenbelow
cos0: = n z =
sin9
m
m
[21]
.
* ---------- .■
Vl H d f/dxf)2 + 0 //d y ') 2
m
M
can be
(6.18)
m
V l+@f fix ')2 + (d fldy')2
It is noted that the back-scattered field is depolarized because in general a 12 does not
vanish. Eq. (6.16) can also be expressed as
i
Sui
+ f j g 2< / S S ) e ™ dx'dy'
Sm
( 6 .20 )
Let’s elaborate the first term of the above equation and rewrite it as
B = j g l ( x ' y j ) e j2kaf dx'dy'
Sm
= j g x(x' y / )ei7k* (n i)ds'
Sm
Define
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(6.21)
- 135-
A '(z') = \g \(x ' ,y\z') ctfdty'
Sm
(6.22)
where A '(z') is the projection of the weighted surface area for z > z' onto the x ' -y '
plane. Eq. (6.21) can then be rewritten as
B = f ex p t-y ^o z') — ~ — dz'
o
°z
(6.23)
By comparing Eq. (6.23) and Eq. (2.40), one notes that the single difference is the
definition of s' (z' ) and A' (z'). The former is the projection of the surface area while
the latter is the projection of the weighted surfacearea.
It has been pointed out that Eq. (2.40) canberewritten as Eq. (2.83). Similarly,
Eq. (6.23) can also be written as
N
;=i
J J d Z L exp0.2M ;.) ( ^
n=i
m
r
'
) . W
a/;-
i )
a/;-
(6.24)
where Z;+ and Z- are points at which discontinuities in the A '(z') or its derivatives
exist.
The reflection coefficient of a metallic object is always -1 for all frequencies.
The range profile of a metallic object renders range information for those points at
which s' (z') and its derivatives are discontinuous. These discontinuities usually give
information about the object’s shape. However, the range profile of an absorbercovered body is more complicated. No simple statement is sufficient to describe the
properties of that range profile because the absorber-covered surface reflection
coefficient is a function of frequency as well as a function of the constitutive parame­
ters of the absorber. However, if we assume that multiple reflections between the
absorber cover and the conductor’s surface can be neglected and the thickness of the
absorber is much smaller than the range resolution of the imaging system, we can
obtain reflection coefficients varying slowly with respect to frequency. In that case
the absorber-covered body’s range profile will have peaks at those ranges where
A '(z') and its derivatives are discontinuous. It is noted that A'(z') can be decreased
by absorber claddings which reduce the surface reflection coefficients. Furthermore,
discontinuities in A'(z') can be artificially created by use of absorbing materials at
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 136 -
unexpected regions of the metallic surface. These created discontinuities may distort
the object’s shape information and deceive visual recognition systems. The above
analysis gives theoretical background for the RCS reduction and image distortion by
absorber covering.
6.2 Radar Absorbing Materials (RAM)
Radar absorbing materials (RAMs) are those materials which can absorb the
incident radar wave energy. RAM was first developed by Germans during the latter
part of World War II to overcome radar detection of their submarines. After 40 years
of development, many different types of RAM have been fabricated. RAMs are used
in a wide variety of applications, examples are: improvement of antenna patterns,
reduction of undesirable reflections from objects and devices, achievement of
reflectionless environments (anechoic chamber), and reduction of radar cross section.
Basically, RAM can be divided into four groups [81]: magnetic absorbers, resonant
absorbers, broadband absorbers, and surface current absorbers. In the following we
will briefly describe each of these.
A. Magnetic Absorber
In magnetic absorbers, the value of the complex magnetic permeability is in the
same range as the complex permittivity. Therefore, the characteristic impedance of
these media is close to that of free space, and reflection from the absorber’s surface
will be small. If the thickness is great enough to attenuate the incident wave to a
significant degree.before it reaches a metallic surface, the overall reflection will be
small. Examples of this group are ferrites. The reflection coefficient is reduced with
these material from 10 to 25 dB in the frequency range of 30 MHz to 1000 MHz
[81].
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 137 -
B. Resonant Absorbers
At microwave frequencies, most materials have higher permittivity than permea­
bility in magnitude.
The characteristic impedance of the materials is therefore
smaller than that of the free space. The difference in characteristic impedance causes
a high reflection from the air / absorbers interface. However, it is possible to cancel
the reflection by suitably introducing another reflection from the interface of the
backside of the absorbers and the object. RAMs have been classically regarded as
resonant when they are a quarter-wavelength thick. However, the actual absorber
thickness used to minimize the reflectance of a material is rarely a quarter
wavelength. If the material is dominantly magnetic in character, the required thick­
ness for minimum reflection may even exceed a half wavelength [83].
C. Broadband Absorbers
The basic concept of broadband absorbers is to devise a material which has
|ir = er at the front surface and has thickness great enough to attenuate the reflection
from the back interface. Ferrite can accomplish this in the VHF and lower UHF
range. The other concept is accomplished by varying the impedance of an absorber
gradually from its incidence surface to a lower impedance at its rear surface. The
dielectric constant is very low at the front surface and increased to very high at the
rear. Therefore, the front face reflection is greatly reduced and the required thickness
is also reduced. Broadband absorbers are divided into two types: the flat-faced type
and the geometrically shaped type. The first type is usually made up of stacked
layers, while the second type achieves the impedance gradient by a geometrical tran­
sition from free space into a lossy media. The geometrically shaped type usually
takes the form of carbon-loaded, flexible foam pyramids or wedges. This is the kind
of absorber widely used in the construction of anechoic chambers.
D. Surface Current Absorbers
In contrast to the three groups above, which are usually designed for reducing
specular reflection or main-lobe reflection, surface current absorbers are designed to
attenuate non-specular or side-lobe energy. Non-specular returns occur when currents
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 138 -
encounter discontinuities, such as gaps, edges, or comers. Surface current absorbers
attenuate the currents before they encounter such discontinuities. Analytical method
for predicting surface current absorber performance are based on the dielectric slab
guided wave analysis [33]. The surface impedance is the key variable in relating the
performance to the material configuration. Surface current absorbers can be in the
form of a thin sheet, or a paint-on version. They usually have a high permeability.
6.3 Experimental Results for Absorber-Covered Bodies
As an aid for presenting our results, some common measurement notations will
be defined first. The transmitting antenna used in the following experiments is a
righthand circularly polarized antenna. The measurement of the scattered field will be
referred to as co-polarized if the receiving antenna is lefthand circularly polarized;
and is referred to as cross-polarized if the receiving antenna is righthand circularly
polarized. The RCS of an object treated in this chapter is either the RCS at a certain
frequency or the mean RCS averaged over a specified spectral band. Three spectral
bands are studied in this chapters. They are : band 1 ( 6 GHz to 11.25 GHz), band 2
(11.25 GHz to 16.5 GHz), and band 3 (6 GHz to 16.5 GHz). The calibration pro­
cedure stated at the end of Sec. 2.6 has been used to obtain the object’s absolute
RCS at each frequency. The RCS of a given conducting sphere is theoretically cal­
culated and experimentally measured. The absolute RCS of the object is then found
through the calibration procedure. The unit of the absolute RCS used in this chapter
is in dB square centimeter ( dB scm), because the conventional unit dB square meter
(dB sm) is too big for the RCS of objects used in this dissertation.
Characteristics of a plate covered with an absorber are of primary interest in this
chapter.
The experimental arrangement of the plate referred in this chapter is
represented in two geometries as shown in Figs. 6.6(a) and 6.6(b). The transmitting
and receiving antennas are both lying in the azimuthal plane. In Fig. 6.6(a), the plate
is perpendicular to the azimuthal plane and two edges of the square plate are always
normal to the line of sight. In a bistatic system we define the line of sight as the
bisector of the angle between the transmitter and receiver (T/R) during the course of
rotation. A zero degree aspect angle is defined as the angle at which the normal of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
139
RECEIVER
TRANSMITTER
(a)
RECEIVER^
TRANSMITTER
b
Fig. 6.6 Geometry of the measurement arrangement of a conducting plate.
(a). Geometry 1— the plate is perpendicular to the azimuthal plane.
Zero degrees is defined as the specular direction.
(b). Geometry 2— the plate is tilted with an angle defined
as the angle between the plate and the rotational axis.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 140 -
the plate coincides with the line of sight In Fig. 6.6(b) the plate has a tilt angle 0,
defined as the angle between the rotational axis and the plate. The arrangements in
Figs. 6.6(a) and 6.6(b) will be referred to as geometry 1 and geometry 2 throughout
this chapter.
In this section, we will discuss some experimental results pertaining to RCS and
images of an object covered with absorbers. The two types of absorbers used are
broadband absorbers and surface current absorbers.
6.3.1 Broadband Absorber-Covered Bodies
The broadband absorber used is the Emerson Eccosorb AN72. AN72 is a light
foam sheet absorber. Typical electrical performance of the Eccosorb AN series is
shown in Fig. 6.7.
TYPICAL ELECTRICAL PERFORMANCE OF ECCOSORB AN
0
8
w
N
F
R
0M
M
E
T
A
L
P
L.
A
T
E
100
FREQUENCY G Ht
• «
lo w e s t
FREQUENCY
f o r i*
m a x im u m
RE FLEC TIV ITY
Fig. 6.7 Typical performance of the Emerson product AN absorbers.
The dielectric constant of a broadband absorber is usually tapered to match the
characteristic impedances of air and conductors. The AN72 absorber has different
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 141 -
surface resistance on the two sides and is an inhomogeneous material. The two sides
of the AN72 are different in color. The front surface is a white-gray color and is
designed to match the characteristic impedance of air, while the back surface is a
bronze color and is designed to match that of a conductor. In our work AN72 was
used to cover a conducting plate (30 cm x 30 cm). The measured co-polarized
attenuation in reflectivity of the metal plate when covered with AN72 in the normal
way (bronze side being attached to the conductor) and in the reversed way (whitegray side being attached to the conductor) in the broadside direction are shown in
Fig. 6.8. The attenuation factor is about 8 to 15 dB in this frequency range when the
plate is covered in the normal way and is much smaller when covered with the
absorber reversed.
i
i
I
V
0.00
E-01
a
covered inside out
r
i
t
u
e
d
i
n
covered in normal way
d
b
b. 'J0E+00
• FPEQUEMCY IN GHZ
Fig. 6.8 Attenuation in dB of a metal plate covered with absorber AN72 in the normal
way and in the reversed way, 0 dB being the response of the plate without absorber
covering.
Next, the square plate is covered with AN72 on both sides. The measurement
arrangement is that of geometry 1. The co-polarized and cross-polarized RCS patterns
of the plate before and after diaphanization at frequency 11.25 GHz are shown in
Figs. 6.9(a) and (b) respectively. The co-polarized and cross-polarized mean RCS
patterns of the plate before and after diaphanization versus rotation angle averaged
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 142-
over band 1 and band 2 are shown in Figs. 6.9(c) and (d) respectively. It is noted
that the co-polarized mean RCS has been reduced a great deal for small rotation
angle and the amount of reduction decreases as the rotation angle increases. The
mean RCS after diaphanization are even greater than those before diaphanization
when the rotation angle is greater than 65°. A possible reason for the above
phenomenon is that AN72 (12 mm in total thickness for two sides) is much thicker
than the thickness of the metallic plate (1.6 mm). The effective area facing to the
T/R after application of the absorber covering is much greater than that of the con­
ducting plate when the plate is rotated close to the edge-on direction. Besides, AN72
is not designed for attenuating side-lobe energy. However, the cross-polarized mean
RCS after diaphanization is much smaller than that before diaphanization for every
aspect angle. This phenomenon is contradictory to the co-polarized case and its rea­
son needs more investigation.
The co-polarized sinograms and cross-polarized sinograms before and after
diaphanization are shown in Figs. 6.9(e) and (f) respectively. For each sinogram
display the bottom line represents the range profile of the first aspect angle while the
top line represents the range profile of the last aspect angle. The range profile is
obtained by Fourier transforming the range-corrected field over band 3. The magni­
tude of the range profile is proportional to the brightness of the display. The sino­
grams are displayed in linear scale. The dynamic display range has been suitably
chosen so that weak signals will not be overriden by strong signals. The maximum
value of the reconstructed sinogram and its dynamic display range are shown in the
attached caption. The rotation angle coverage is also noted in the Figures. From the
sinograms illustrated, one can conclude that the main contributors to the scattered
field of a conducting plate arranged in geometry 1 are the two vertical edges. After
diaphanization the equivalent scattering strength of the rear edge (the one farther
away from the T/R) has been reduced a great deal for every rotation angle as can be
seen from the reduced brightness in the sinogram. However, the co-polarized scatter­
ing strength of the front edge (the one closer to the T/R) has been increased when the
plate is rotated close to the end-on direction. The co-polarized images reconstructed
over an angular window from <f>= 0° to 90° before and after diaphanization are shown
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 143 J
without absorber
9 .i t
E -il
v jv y
vV1 *
with absorber AN72
t.ttE -tl
2 .2 !E -il
—
rotation ancle in decree
(a)
S .tt
E«tl
3.78
£♦•1
2. St
E -tt
U
without absorber
k
................... ..
■
with absorber AN72
1.25
E « tl
v
f
t
t
¥
f
\
i
ROTATION ANCLE IN DECREE
(b)
s .t t
E - il
conducting plate, co-polarized
with AN72, co-polarized
. . . . . . . conducting plate, cx-polarize
with AN72, cx-polarized
3 .St
£-tl
2. t t
E -tt
• ‘A.1.........?.*!
S .t t
E’ t t
9 .0 9 E -tl
2 .2SE-91
A .S tE -tl
—
ro ta tio n ancle :n decree
6 ..'1 E -il
(c)
t.t*
E-tl
3. St
E -tt
conducting plate, co-polarized
with AN72. co-polarized
* conducting plate, cx-polarized
—
with AN72, cx-polarized
i
‘V
2 .tt
E -t l I
A ...
w fi
S .tt
E -tt
t»,
i.itE -9 1
2.2SE-tt
».!♦£.#1
ROTATION ancle in oecree
C.75E-9I
(d)
F ig . 6 .9 (cap tion se e page 145)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
AN72 cowered p l a t e ; co-p o lo r
max:18; d ls p ( 1 .0 ) ; (0.90)deg
e
co nductin g pi a t e ; c x - p o l a r ; (0,90)
—max—13;d 1s p (1 ,0 )
p l a t e w ith A N 72;cx-polar;(0,90)
max=2.2; d l s p ( 1 ,0 )
f
v e r t i c a l c o n d u ctin g p l a t e
• a x : 89; d l» p (8 5 » G 5 );(0 » 9 0 )deg
. v e r t i c a l PK72 co v ered p la te
max: 83; d lip < 0 6 .6 5 ); (0 .9 0 )d e g
g
Fig. 6.9 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 145 v e r t ic a l p la te ;
( 3 0 .9 0 ) d e g
v e t le a l PN72 c o v e re d ;
max: 7 9 ; d ls p ( 7 5 .5 5 )
max: 7 8 ; d ls p ( 7 5 . 5 5 )
n
c o n d u c tin g p l a t e ; c x - p o l a r
•o x :‘7 2 ; d l s p ( 7 0 .5 0 ) ; (3 0 ,9 0 ) d e g
PN72 c o v e re d p l a t e '; c x - p o l a r
max 5 9 , d ls p ( 7 0 . 5 0 ) ; ( 3 0 , 9 0 ) d e g
Fig. 6.9 Measured RCS pattern, sinograms and reconstructed images of a
conducting plate before and after diaphanization with geometry 1.
(a). Co-polarized RCS pattern at frequency 11.25 GHz.
(b). Cross-polarized RCS pattern at frequency 11.25 GHz.
(c). Co-polarized and cross-polarized mean RCS patterns averaged over band 1.
(d). Co-polarized and cross-polarized mean RCS patterns averaged over band 2.
(e). Co-polarized sinograms of the plate before (above) and
after (below) diaphanization.
(f). Cross-polarized sinograms of the plate before (above) and
after (below) diaphanization.
(g). Co-polarized images of the plate reconstructed over an angular window
from <|>= 0° to 90° before (left) and after (right) diaphanization.
(h). Co-polarized images of the plate reconstructed over an angular window
from <[)= 30° to 90° before (left) and after (right) diaphanization.
(i). Cross-polarized images of the plate reconstructed over an angular window
from <J)= 30° to 90° before (left) and after (right) diaphanization.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 146 -
in Fig. 6.9(g). The brightness of the images is displayed in log scale. The maximum
value and the dynamic display range of the images are also noted in the accompany­
ing captions of these figures. The co-polarized and cross-polarized images recon­
structed over an angular window from 0=30° to 90° before and after diaphanization
are shown in Figs. 6.9(h) and (i) respectively. The difference in appearance between
the images reconstructed over different angular windows can be explained by backprojecting the sinogram into the image space as described in Chap. 3. It is also
noted that the brightness in images has been reduced after diaphanization.
Next, we used AN72 to cover a complex shaped object. A scale model B-S2
airplane is partly covered with absorber. The covered portions have been shown in
Fig. 5.3. The co-polarized and cross-polarized mean RCS patterns averaged over
band 3, sinograms, and images of the scale model B-52 before and after diaphaniza­
tion are shown from Figs. 6.10(a) to Fig. 6.10(f).
The above results show that the broadband absorber is effective at attenuating
main-lobe energy but not effective at reducing side-lobe energy. In the next subsec­
tion we will use an alternative absorbing material to cover a metallic object.
6.3.2 Surface C urrent Absorber-Covered Bodies
The surface current absorber used is the Emerson product Eccosorb GDS, a thin
(0.8 mm in thickness), high ioss, silicone rubber sheet that can be bonded to curved
surfaces. The characteristics of GDS are shown in Table. 6.1. GDS will lower the
reflection by attenuating surface waves and multiple bounces.
GDS is used to cover a conducting plate (30 cm x 30 cm). Two covering
configurations are tested. The first configuration covers the whole plate by GDS on
both sides as shown in Fig. 6.11(a). The second configuration covers only the verti­
cal edges as shown in Fig. 6.11(b). It has been found that the induced surface
current density of a conducting strip when illuminated by a plane wave is highest in
the neighborhood of an edge [77]. Covering only the edges with GDS is therefore
expected to reduce most of the return energy. When the measurement is arranged in
geometry 1 with the covered edges parallel to the rotational axis, the co-polarized
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 147 -
4.
3.
2
1.
'> line number
comparison o f b 5 2 r l . p and b 5 2 r l . c ! 0 ; a v e .3
I
I
V
a
n
fi
t
e+91
u
d
e
i
n
d
b
l.OOO0e+00
> lina number
com pariso n o f b S 2 r r . p and b 5 2 r r . c ! 0 ; a v e . 3
b 5 2 r 1 . p ; m a x : 4 1 ; d l s p ( 3 . 0 ) ; ( 0 ,9 0 ) d e g
n
b 5 2 r r . p ; m a x > 7 ; d l» p ( 3 .0 ) j( 0 » 9 0 ) d e g
r. '
1
b 5 2 r 1 . c ;m a x : 5 ; d l s p ( 3 . 0 ) ; ( 0 . 9 0 ) d e g
f- A\:-
c
r
b S 2 rr.c .rn ax :2 .7 ;d i» p (3 .0 );(0 .9 0 )d
d
Fig. 6.10 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-1 4 8 -
b S a^-J.p ; w ith o u t a b t o r b e r
max 8 9 ; d 1« p ( 0 4 ,6 0 ) ; ( 0 , 9 0 )d e g
bSZrrt.pi c x - p o la r ;
.
• a x <63; d l » p ( 9 4 . 6 0 ) ; (9 « 9 0 )d e g
b52M c ; w ith o b to r b e r .
max BS, d l* p (B 4 .G 0 ); C 0 ,9 0 )d eg
b S 2 r r .c ; w ith a b s o r b e r ) c x -p o la r
max ^ ;
d l t p ( 9 4 , G 0 ) ; ( 0 ,9 O ) d e g
F ig . 6 .1 0 M e a n R C S p a tte rn s, sin o g ra m s, a n d re c o n stru c te d im a g es o f the
B -5 2 a irp lan e b e fo re a n d a fte r d iap h a n iz a tio n .
(a). C o -p o la riz e d m e a n R C S p a tte rn s a v era g ed o v e r b an d 3.
(b). C ro ss-p o la riz e d m ea n R C S p a tte rn s a v era g ed o v e r b a n d 3.
(c). C o -p o la riz e d sin o g ra m s b efo re (ab o v e) a n d a fte r (b elo w ) d iap h a n iz a tio n .
(d). C ro ss-p o la rize d sin o g ra m s b e fo re (ab o v e) a n d a fte r (b e lo w ) d iap h a n iz a tio n .
(e). C o -p o la riz e d im a g e b e fo re (left) an d a fte r (rig h t) d ia p h a n iz a tio n
re c o n stru c te d o v e r an a n g u la r w in d o w fro m 0 = 0 ° to 9 0 °.
(f). C ro ss-p o la riz e d im a g e b e fo re (left) and a fte r (rig h t) d ia p h a n iz a tio n
re c o n stru c te d o v e r an a n g u la r w in d o w fro m (|)=0o to 9 0 °.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 149-
Table 6.1 Characteristics of the GDS absorbers
Frequency (GHz)
3
8.6
Dielectric const k'
20
20
Dielectric loss tangent tanS
0.67
0
Magnetic permeability Km'
3.5
1.2
Magnetic loss tangent tanSm
0.4
0.4
Attenuation dB/cm
Relative Impedance IZ l/Z0 l
1.1
0.4
45
0.4
and cross-polarized mean RCS patterns of the plate without and with GDS covering
averaged over band 1 and band 2 are measured and shown from Fig. 6.11(c) to Fig.
6.11(f). It is noted that the mean RCS has been reduced after applying a GDS cover­
ing, and the amount of reduction increases as the rotation angle increases. This
confirms the effectiveness of GDS at reducing side-lobe energy or surface waves.
The amount of reduction over band 1 is smaller than that over band 2, implying GDS
is more effective at higher operating frequencies. One can also find that covering
edges is not as effective at reducing RCS as covering the whole plate. An explana­
tion for this observation can be obtained via an examination of the range profiles
which render information on scattering mechanisms.
The co-polarized and cross-polarized sinograms of the GDS-covered plates are
shown in Figs. 6.11(g) and (h) respectively. By examining the co-polarized sino­
grams of Fig. 6.11(g), one notes two more traces in the sinogram of the second
configuration. These extra traces are reflected from the boundaries of the absorber
and the plate.
Surface impedance is discontinuous at these boundaries.
This
phenomenon verifies the statement that reflections occur at those regions with discon­
tinuous surface impedance or reflectivity. However, the cross-polarized sinogram of
the second configuration as shown in the lower part of Fig. 6.11(h) has a different
appearance. No extra trace is found in the front edge. This phenomena needs more
investigation. The co-polarized and cross-polarized projection images of these two
GDS-covered plates are shown in Figs. 6.11(i) and (j) respectively. It can be seen
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 150-
GDS
GD S
aluminum plate
*■
aluminum plate
(a)
;
j
V
a
*
?
i
t
Q
d
a
i
n
<b)
S . 90
E+01
conducting plate
edge covered w ith GD S
whole covered with GDS
3 .9 0
E+01
2.00
E+01
5 .00
E+00
d
b
♦ . 50E+01
. 00E-01
2.25E+01
— ROTATION ANGLE IN DEGREE
6 . 75E+01
(c)
5 .0 9
E+01
conducting plate
edge covered with GDS
whole covered with GDS
ROTATION ANGLE IN DEGREE
(d)
Fig. 6.11 (caption see page 152)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 151 -
conducting plate
***** edge covered with G D S
whole covered with GDS
3
It
d
E+00
2.25E+01
ROTATION ANGLE IN DEGREE
(e)
r
conducting plate
******* edge covered with GDS
whole covered with GD S
V
-V
V
______________ ________
E+00
-J, 00E-01
2.25E+01
4.S0E+01
- - > ROTATION ANGLE IN DEGREE
6.75E+01
(0
GDS uihole cowered; c x -p o la r- ;
m a x : 4 .8 ; d i s p ( l < 0 ) ; (0 ,9 0 ) d e g
‘
GDS e d g e co w ered ; c o - p o l a r
max : 6 . 9 ; d l t p ( l . O ) r ( 0 , 9 0 ) d e g
Fig. 6.11 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fig. 6.11 Covering configurations of a plate and their mean RCS patterns, range
profiles and reconstructed images, (a). Both sides are covered with GDS. (b). Only a
pair of edges are covered with GDS. (c). The mean co-polarized RCS averaged over
band 1. (d). The mean co-polarized RCS averaged over band 2. (e). The mean cross­
polarized RCS averaged over band 1. (f). The mean cross-polarized RCS averaged
over band 2. (g). Co-polarized sinograms of the GDS wholly covered plate (above)
and edge covered plate (below), (h). Cross-polarized sinograms of the GDS wholly
covered plate (above) and edge covered plate (below), (i) Co-polarized images of the
GDS wholly covered plate (left) and edge covered plate (right) reconstructed from an
angular window from <J)=30° and 90°. (j) Cross-polarized images of the GDS wholly
covered plate (left) and edge covered plate (right) reconstructed from an angular win­
dow from <t>=30° and 90°.
with permission of the copyright owner. Further reproduction prohibited without permission.
- 153-
that brightness of the rear edge has been reduced as compared to the left images of
Figs. 6.9(h) and 6.9(i), which are the images of the same plate without absorber cov­
ering. This fact implies that diffraction from the rear edge has been suppressed
because the brightness of an image pixel is proportional to the mean scattering
strength of that portion averaged over the specified angular window and spectral win­
dow as stated earlier. If we compare the RCS plots and the reconstructed images, we
can conclude that covering the whole plate is more desirable from the detection
avoidance point of view. However, covering the edge portions is more effective at
distorting the image because the discontinuities in the edge covered piate give rise to
twin projection images at each edge, distorting thereby the image as compared to that
of the wholy covered plate.
The second object tested was an aluminum tube with one end covered by a cap
and the other end open. Its geometry and dimensions are shown in Fig. 6.12(a). The
motivation for studying this object stems from its likeness to an engine intake. The
tube is placed with its axis perpendicular to the rotational axis. The direction normal
to the cap is defined as zero degrees. The co-polarized mean RCS pattern of the tube
averaged over band 3 is shown in the solid curve of Fig. 6.12(b). Its sinogram is
shown in the upper part of Fig. 6.12(c). Examining the sinogram, one can find that
multiple reflections are important contributors to the back-scattered field when the
rotation angle exceeds 110°, because extra peaks can clearly be seen beyond the
maximum dimension of the tube. Besides, the range distance of these extra peaks
decreases as the rotation angle increases, which can be explained by the plot of the
ray paths as shown in Fig. 6.12(d). The incident ray can be reflected to the receiver
in the following ways: (1). It can impinge on the rim of the empty mouth and then
be diffracted to the receiver. (2). At a later time it may hit the rim of the back cap
and be diffracted. (3). It may enter into the mouth, bouncing between the walls, be
reflected back by the cap, and finally emerge from the mouth and propagate directly
to the receiver. (4). The ray may reflect back from the cap, impinge on the rim of the
mouth, and be diffracted to the receiver. The ray path lengths of cases (3) and (4)
decrease as the rotation angle is increased, which explains the appearance of the sino­
gram.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
alum inum tube
r
7.5 c m
capped
—* !4 c m |* -
~ H 4cm h -
30 cm
(a)
receiver
<t>= 0°
<t> = 180'
transm itter
<)> = 90'
w ith ou t absorber
w ith absorber
♦ . 3OE»01
- ■ ROTATION ANCLE IN DEGREE
(b)
F ig. 6 .1 2 (cap tio n see n ex t p ag e)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 155t u b e ; ~'(Q, 180)
max:14; d l s p (1,0)
tub e ; with abs or be r ; ( 0 ,i© 0 )
max: S . 7 ; d l s p ( 1 , 0 )
r ••
i
d
tu b e ;d = 3 " , L = 1 2 "; ( 1 0 0 . 1 8 0 )d eg
max =0$<.^dj*p ( 7 5 .5 0 ) .
a b t o r b e r c o v e re d . t u b * ; ( 1 0 0 .1 0 0 )
max =6 9 ; d l s p ( 6 5 .5 0 )
e
Fig. 6.12 Geometry, mean RCS, sinogram, and reconstructed image of a tube. (a).
Geometry of a tube and the sketch of the absorber covered regions, (b). Co-polarized
mean RCS patterns before and after absorber covering, (c). Sinograms before (above)
and after (below) diaphanization. (d). Ray tracing of an incident wave travelling inside
a hollow tube with a smaller rotation angle (left) and larger rotation angle(right). (e).
Images before (left) and after (right) diaphanization reconstructed over an angular win­
dow from <t>=100° to 180°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 156-
Assume the angular window of interest is from <(>=100o to 180°. Multiple
reflections and rim diffractions are the dominant scattering mechanism over this win­
dow. To reduce multiple reflections and edge diffractions, we cover both sides of the
cap with an AN72 broadband absorber, and cover the inner and outer rings of the
mouth and the outer ring of the other end with a GDS absorber. The covered pat­
terns are also sketched in Fig. 6.12(a). The mean RCS pattern after application of
the absorber covering is shown in the dashed curve of Fig. 6.12(b). It is seen that
the RCS after diaphanization has been reduced noticeably over the specified window.
The sinogram after diaphanization is shown in the lower part of Fig. 6.12(c). It is
seen that multiple reflections have been suppressed and the edge diffraction strength
has also been reduced.
The co-polarized images of the tube before and after
diaphanization are shown in Fig. 6.12(e).
6.4 Target Shaping
The purpose of shaping is to manipulate or deform the target surface so as to
reflect the incident energy in directions away from radar. However, a reduction in
RCS at one viewing angle is usually accompanied by an enchancement at another
angle when the surface is reshaped. The success of shaping depends on the existence
of angular sectors over which low RCS is less important.
Shaping is a trade-off between structural and electromagnetics requirement. The
general procedure of shaping a target is as follows [33]:
1.
Define the threat sectors, i.e., the solid angle relative to the target over which
RCS is to be reduced.
2.
Take advantage of shaping without resorting to absorbing materials — Shape
must be chosen to optimize the mission of the vehicle, and the mission itself
determine the blend of RCS reduction versus structural characteristic. This
requires iterations of RCS and aerodynamic performance predictions for a set of
shapes and then verification by testing.
3.
If the final configuration cannot meet the requirement, use absorber cladding
provided this is aerodynamically permissible.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 157 -
To study the effect of target shaping on the RCS, it is helpful to examine the
dependence of the various analytic components of the RCS on frequency. At high
frequencies, the RCS is dependent on the surface curvature of the equi-phase points
if specular reflections are involved (see Sec. 2.4.1). The RCS is proportional to f 2
for a flat surface; proportional to / for a single-curved surface; and independent of /
for a double-curved surface. For example, the specular RCS of a plate with dimen­
sion w x l , of a cylinder with radius a and length /, and of a sphere with radius a are
as follows:
a =
47t/2W2
X2
plate
2ttal2
X
cylinder
7ia2
(6.25)
sphere
When one of the radii of the curvaturegoes to zero, an edge is then created.
The diffraction fields of an edge with
length / can be written as follows (see Sec.
2.3.3):
E d = j k D i k f i j i))/sny p ° se)
kl cosQ
(6.26)
where D is the diffraction coefficient, whose expression has been shown in Eq.
(2.44), and D is proportional to 1I k. If the inclination angle is 0=90°, i.e., normal
incidence to the edge, then sinc(£/cos0) = 1. Therefore, E d is independent of k and
is proportional to /. In the side-lobe region, for a given 0, E d is proportional to
sinf/WcosQt
e jk lc o sQ
-jk lc o s Q
— ----«— =
•
- ------, the equivalent diffraction of the end points (or tips)
xcos0
2jk cos0
e ± jk lco sB
}
is proportional to I------------1 = ----- -— —. Therefore, the RCS of a tip is propor2jkcosQ
2y£cos0
tional to 1/A:2 and is independent of /. In the case of surface discontinuities, the fre­
quency dependence of the n th surface derivative discontinuities is k~2(-n~l\ as seen in
Eq. (2.83). In short, the RCS of a body is dependent on the operation of several
kinds of mechanisms. These mechanisms can be categorized according to their
strength and can be listed in hierarchy according to their dependence on the fre­
quency and dimension. Shown in Table. 6.2 is the hierarchy of scattering shapes
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 158 -
Table 6.2 Hierarchy of Scattering Shapes [52]
G eom etry
Type
Freq.
Dep.
Size
Dep.
Formula
M aximum
square
trihedral
com er
retro-reflector
right
dihedral
com er
reflector
a = 12jca4/X2
M aximum
o = 8ita2b2/X2
m axim um
Flat plate
a = 4ica2b2/\2
norm al incidence
m axim um
C ylin d er
norm al incidence
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 159 -
Table 6.2 Hierarchy of Scattering Shapes (continued) [52]
Freq. Size
Type
Geom etry
Formula
Dep. Dep.
sphere
F°
L
2
m axim um
a=
jta 2
norm al incidence
Straight
edge
normal
incidence
k
/(9 .
F°
V
L 2
L 2
0 - aspect
K
Curved
edge
normal
incidence
/ ( 6 . e te) a l / 2
F -i
L
&
1
0,/u * interior dihedral
angle between faces
meeting at edge
X2 g (a
Apex
F -2
L°
a > \
,(3,0,0)
. P - interior angles
of tip
0, 0 - aspect angles
a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 160 -
[52]. This table is very helpful, because it provides a fundamental rule for shaping a
target. It is obvious that in order for the RCS to become smaller with increasing fre­
quency, no specular contribution to the scattered field should exist. The body should
appear as a tip or an edge at the aspect at which the RCS is to be minimized.
It is interesting to note that the frequency dependence of the RCS of a rectangu­
lar plate is affected by the aspect angle. In the specular direction, the RCS is propor­
tional to &2 and the square of the area; at the aspect angle with incident wave normal
to a pair of edges, the RCS is independent of frequency but is proportional to the
square of the length of that edge; at side-lobe regions, the RCS is proportional to
1/k2 and is independent of the edge length.
As discussed earlier, a reduction in RCS over a range of aspect angles is usually
accompanied by an enchancement in another range of angles when the surface is
reshaped. Some examples will be given next to verify the above statement
A square conducting plate (40 cm x 40 cm) is reshaped into a serrated square
plate by serrating its edges as shown in Fig. 6.13(a). The scattering properties of the
serrated plate and the unserrated plate which we will simply refer to as the plate will
be studied and compared. First, we use geometry 1 to measure the scattered field.
In this arrangement the line of sight is normal to the two vertical edges of the plate
but not perpendicular to each serrated edge. Therefore these two vertical edges are
the dominant contributors to the back-scattered field of the flat plate, while the tips
are the major contributors to the scattered field of the serrated plate. The mean co­
polarized RCS of the plate and serrated plate averaged over band 1 and band 2
versus rotation angle are shown in Fig. 6.13(b) and Fig. 6.13(c) respectively. Com­
paring Figs. 6.13(b) and (c), one can find that the mean RCS patterns of the two
bands differ very little for the flat plate case except that the mainbeam width is more
narrow in band 2. However, the mean RCS patterns of the serrated plate are greater
in band 1 than in band 2. The above results are consistent with what was predicted
in Table. 6.2., which shows the scattered field from an edge with normal incidence is
proportional to the edge length but independent of frequency while the contribution
from a tip is independent of edge length but inversely proportional to frequency. The
sinograms and images of the plate and the serrated plate are shown in Figs. 6.13(d)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 161 -
T
—•1 6cm )•—
(a)
flat plate
serrated plate
a
n
?
i
t
•d
u
e
i
n
d
D
a .eeE -at
DOTATION ANGLE
in degree
(b)
flat plate
serrated plate
v
a
«
P
1
i
t
u
d
9
I
n
d
b
ROTATION ANGLE IN DEGREE
(c)
Fig. 6.13 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
v e r tic a l serg ated p la te ;
max:55; d l s p ( 1 , 0 ) ; (0 ,9 0 ) d e g
d
■
E
e
Fig. 6.13 Geometry, and mean RCS patterns of a serrated plate using geometry 1. (a).
Geometry of a serrated plate, (b). Mean co-polarized RCS patterns of a flat plate and
a serrated plate averaged over band 1. (c). Mean co-polarized RCS patterns of a flat
plate and a serrated plate averaged over band 2. (d). Sinogram of the flat plate (above)
and serrated plate (below), (e). images of the flat plate (left) and the serrated plate
(right) reconstructed from data collected over an angular window from <j)=30° to 90°.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 163 -
and (e) respectively.
Next, the measurement is arranged in geometry 2 with a tilted angle 0=90°.
Zero degrees is defined as the angle at which a pair of edges the plate are normal to
the bisector. The co-polarized mean RCS patterns of these two plates averaged over
band 3 are shown in Fig. 6.14(a). Away from the broadside, the major contributors to
the back-scattered field are the tip points. There are more tips in the serrated plate
than in the flat plat Additionally, part of the serrated edges are normal to the bisec­
tor at <(>=450 and 0=135°. These facts explain why the mean RCSs of the serrated
plate are higher than those of the fiat plate in most aspects and why two peaks appear
in the dashed curves at 0=45° and 135°. The mean RCS patterns shown in Fig. 6.13
and Fig. 6.14 show evidence of the fundamental concept of target shaping, that is, a
reduction in RCS at one viewing angle is usually accompanied by an enhancement at
another viewing angle.
The co-polarized sinogram and image reconstructed over an angular window
from 0= 0° to 360° for a serrated plate mounted as in geometry 2 with a tilted angle
0= 60° are shown in Figs. 6.14(b) and (c) respectively. In this experimental arrange­
ment the field scattered from the styrofoam support cannot be subtracted effectively.
The two straight vertical lines in the sinogram and the blurred part at the center of
the image are both from the styrofoam supporter. Therefore we dot not compare the
mean RCS patterns with this measurement arrangement.
In the following example we will study the effect of a sharp tip and a round
comer on the RCS. The objects are a triangular plate with sharp tips and a triangular
plate with two round comers. These two structures are shown in Figs. 6.15(a) and
(b) respectively. The measurement is arranged in geometry 2 with a tilted angle
0=90°, i.e., the plates are lying in the azimuthal plane. 90 degrees is defined as the
nose-on angle. These structures are used to crudely simulate a delta wing of an air­
plane. The co-polarized mean RCS and sinograms of these two plates averaged over
band 3 are shown in Figs. 6.15(c) and (d) respectively. By comparing the traces in
the sinograms of Fig. 6.15(d), one can find that the energy reflected from a round
comer is much less than that from a sharp tip when the edge is not normal to the
bisector. The sinograms also illustrate some other scattering mechanisms which
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 164-
21
v
E+Ol
flat p late
3
M
serrated p late
fi
t
a
d
a
E-01
i
n
d
b
1.35E+02
ROTATION ANGLE
(a )
dag
t i l t e d e e r g a te d p l a t a i 3 9
m a x i GSi d la p ( £ 5 . 4 5 ) 4-
V
/
S N\ VV
/
’
c
Fig. 6.14 Mean co-polarized RCS and images of a flat plate and a serrated plate using
geometry 2. (a). Co-polarized mean RCS a flat plate and serrated plate averaged over
band 3. (b) Sinogram of the serrated plate with a tilted angle 0=60° and an angular
window from 0° to 360°. (c). Image of the serrated plate reconstructed from <))=0o to
360°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 165 -
require more investigation.
A good example of target shaping is the B-l bomber, which appears to be
designed to have little specular reflection over a range of viewing angle relevant to
practical encounters with probing monostatic radar system. A metalized 1:100 scale
model of the B-l was mounted on a rotating pedestal with inclination angle 20° as
shown in Fig. 6.16(a). The origin of the azimuthal rotation angle $ is defined as the
broadside direction and a 90° rotation angle is in the head-on direction. The mean
RCSs averaged over band 3 using co-polarized and cross-polarized waves versus the
rotation angle are shown in Fig. 6.16(b). The sinograms and projection images recon­
structed over an angular window from <j)=0° to 90° of these two polarizations are
shown in Figs. 6.16(c) and (d) respectively.
For this object structure and measurement arrangement, multiple reflections
between the flat bottom and the fuel tanks are important contributors to the RCS for
those aspects free of strong specular reflections. The difference between the copolarized and cross-polarized mean RCS is small for many aspects.
This
phenomenon is different from that of seen with the B-52, where the co-polarized
mean RCS is much stronger than the cross-polarized mean RCS in most aspects.
The effect of multiple reflections on the images generated is the blurring in the image
of the bottom structure as shown in Fig. 6.16(d).
6.5 Combination of Absorber Covering and Shaping
If the final shape of an object is determined and its RCS cannot meet the
requirement, then one should resort to absorber covering. When an absorber is coated
on a metallic object, the boundaries of the covered regions will introduce discontinui­
ties in surface impedance and cause radiations. Scattering hierarchy also exists in the
shape of the coating boundaries. The total scattered field may be more effectively
reduced if the coating boundaries are suitably shaped or tapered.
In the optical region one can paint an object’s surface different colors and pat­
terns so that its appearance differs from the original one. This process is called
camouflage. In the microwave region, if the surface of the object is coated with
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 16630cm
-<
VO
(b)
r ,(a)
w ith sharp tip
w ith round c o m e r s
1.35E+02
ROTATION ANGLE IN DEGREE
(c)
-
— —
l J.
:
t r i a n g u l a r p l a t e ; ro u n d c o r n e r
max: 3 . 0 ; d l a p C . 1 . 0 ) ; ( 0 . 100)deg
(d)
Fig. 6.15 Geometry, mean RCS patterns, and sinograms of triangular plates with sharp
tips and rounded tips arranged in geometry 2. (a). Geometry of a triangular plate with
sharp tips. (b). Geometry of a triangular plate with round comers, (c). Co-polarized
mean RCS patterns of the two plates averaged over band 3. (d). Sinograms of the
plates with sharp tips (above) and rounded tips (below).
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
I
V
co-polarized
a
M
e+01
T
i
******* cx-polarized
t
u
d
e
i
n
E+01
d
b
0 .0 0 E -0 1
2.26E+01
> ROTATION ANGLE IN DEGREE
6.77E+01
(b)
b i u r l ; amx: 1 3 ; d i s p ( 1 , 0 ) ; ( 0 , 9 0 ) d e g
' . JB S fr
b l h r l ; c © - p o la r ; (O ,9 0 )d e g
a a x -8 0 i d l* p C 0 5 /G 0 )
b lu rr;
b l u r r ; c x - p o l a r s ( 0 .9 0 ) d e g
mox^ 6 6 ; d l« p ( 9 5 .G O
m ax :8 ;d l» p ( l > 0 ) ; (0 ,9 0 )d e g
(d)
Fig. 6.16 Picture, mean RCS patterns, and images of a scale model of B-l airplane,
(a). Picture of a scale model B-l airplane mounted on a rotating pedestal, (b). Mean
co-polarized and cross-polarized RCS patterns averaged over band 3. (c). Co-polarized
(above) and cross-polarized (below) sinograms over an angular window from <j>=0° to
90°. (d). Co-polarized (left) and cross-polarized (right) images reconstructed from data
collected over angular window <j)=0° to 90°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 168 -
absorbing materials in different thicknesses or patterns, its microwave images may
also be altered and may become unrecognizable. If the absorber covering is con­
sidered with shaping of the covering boundaries, the technique is referred to as a
combination of absorber covering and shaping.
We will give two examples based on the above considerations. A square con­
ducting plate (30 cm x 30 cm) is covered with GDS and the covering pattern is
shown in Fig. 6.17(a). The measurement is arranged in geometry 2 with a tilted
angle 6=90°. Zero degrees is defined as the angle at which a pair of edges are nor­
mal to the line of sight We shape the covering absorbers in a curved manner in
order to avoid any specular reflections which might be generated by an edge normal
to the line of sight. The co-polarized mean RCS pattern of the plate with and
without an absorber covering are shown in Fig. 6.17(b). It can be seen that the RCS
has been reduced for most aspect angles. Their sinograms over an angular window
from <]>=0o to 360° are shown in Figs. 6.17(c) and (d). One can find that the
equivalent scattering strength of the edges and tips has been effectively suppressed.
The reconstructed images of these two plates are shown in Fig. 6.17(e), where both
images have the same dynamic display range. The brightness of the edges and tips
after diaphanization has been reduced. If the display dynamic range of this coated
plate’s image is lowered, the resultant image is shown in Fig. 6.17(f).
The next example is the scale model B-l covered with a blend of GDS and
AN72. The picture of the absorbor-covered scale model B-l is shown in Fig.
6.18(a). We cover those regions which might cause specular reflections with broad­
band absorbers, and cover the edges and comers with GDS to reduce edge
diffractions and multiple reflections. The measured mean co-polarized and crosspolarized RCS averaged over band 3 before and after absorber covering are shown in
Figs. 6.18(b) and (c). It is seen that the amount of RCS reduction is not great in most
aspect angles for the co-polarized waves, while the amount of reduction is pro­
nounced for the cross-polarized waves. The co-polarized and cross-polarized sino­
grams before and after absorber covering are shown in Figs. 6.18(d) and (e) respec­
tively. Comparing the sinograms one can find that the absorbers can suppress cross­
polarized waves more effectively, which is in agreement with the plots of Figs.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 169 -
firDA
i
V
w ith o u t absorber
;- w ith ab sorb er G D S
a
M
?i
t
u
d
e
i
n
d
b
0 . 00 E-01
•— > ROTATION ANGLE IN DEGREE
(b)
f l a t p la t e ; th eta= 9 0
m a x : 0 . 9 4 ; d l s p ( . 0 S . 0 ) ; ( 0 . 360)deg
GDS c o a t e d p l a t e ; th e t a = 9 0
m a x : 0 . 3; d i « p ( , 0 5 , 0 ) ; (0 ,3 S 0 )d e g
•
t i l t e d p l a t e ; th e ta = 9 0
m ax:S 3; d l t p (S3* 3 0 ) '
t i l t e d c o a te d p l a t e ; t h e ta » 9 0
m ax:4 9 ; d l t p (53 * 3 6 )
-rrr'
I
c o a te d t i l t e d p l a t e ; t h e t a - 9 0
max: 4 9 ; d l t p ( 4 9 ,3 0 )
F ig . 6 .1 7 C o v e re d p a ttern s, m ean R C S p a ttern s an d re c o n stru c te d im a g e o f a p late
arra n g e d in g e o m e try 2 w ith tilted an g le 9 = 90°. (a) C o v ered p a tte rn s o f a plate,
(b ) C o -p o larize d m ean R C S p a tte rn s w ith o u t an d w ith a b so rb e r c o v e rin g av erag ed
o v e r b an d 3. (c) S in o g ra m o f th e c o n d u ctin g p late o v e r an a n g u la r w in d o w fro m
<j>=0° to 3 6 0 °. (d) S in o g ra m o f th e a b so rb er-c o v ere d p la te o v e r a n a n g u la r w in d o w
fro m <|)=0o to 360°. (e). Im ag e s o f th e p la te b e fo re (left) an d a fte r (rig h t) a b so rb er
c o v erin g . B o th im a g es h a v e the sa m e d y n a m ic d isp lay ran g e, (f). Im ag e o f the
a b so rb e r-c o v e re d p late w ith a lo w e re d d y n a m ic d isp lay ran g e.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 170-
(a)
w ith o u t ab so rb er
w ith ab sorb er
0 . 00 E-01
2.26E+01
— • ROTATION ANGLE IN DEGREE
6. 77E+01
(b )
—
a
■+*++ w ith ab sorb er
n
T
i
t
u
d
w ith o u t ab sorb er
E+01
E+01
a
i
n
E+01
d
b
0OL-01
->
ROTATION ANGLE IN DEGREE
(C)
Fig. 6.18 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
b l u r m ; w ith o u t a b t o r b e r
• a x - 6 2 ; d l t p ( 0 6 ,9 6 )
b l u r r c ; a b t o r b e r c o v e re d
max:
d i t p (6 0 ,5 6 )
'4 # ~
"rLW v i \
Fig. 6.18 Pictures of the absorber-covered scale model B-l and its RCS and recon­
structed images before and after diaphanization. (a) Pictures of the absorber-covered
scale model B -l, top view (left) and bottom view (right), (b) Mean co-polarized RCS
patterns before and after diaphanization averaged over band 3. (c) Mean cross­
polarized RCS patterns before and after diaphanization averaged over band 3. (d) Copolarized sinograms of B-l before (above) and after (below) diaphanization. (e).
Cross-polarized sinograms of B-l before (above) and after (below) diaphanization. (f).
Co-polarized images of B-l before (left) and after (right) diaphanization. (g). Cross­
polarized images of B-l before (left) and after (right) diaphanization.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 172 -
6.9(c), 6.9(d), 6.11(e), and 6.11(f)- The co-polarized and cross-polarized images
before and after absorber covering are shown in Figs. 6.18(f) and (g) respectively.
By examining Fig. 6.18(g), one can see that multiple reflections have been
suppressed for the cross-polarized waves. However, the images are still recognizable.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 173 -
CHAPTER 7
DIAPHANIZATION BY IMPEDANCE LOADING
Results in the previous chapter confirm that covering an object with absorbers
reduces the RCS and altering the object shape can direct the field away from a
searching radar. Images can somewhat be distorted by the introduction of artificial
discontinuities in the surface impedance of an object. However, the shape of a com­
plex object can still be recognized even when the scattering strength of the hot spots
has been reduced. In this chapter we will investigate another approach, impedance
loading, and examine its effect on RCS and images.
The scattered field from a conducting body with impedance loading can be
expressed as the sum of a component independent of the load value and a loaddependent component [21]. Accordingly, it is possible to specify the impedance so
as to minimize the back-scattered field for a particular frequency and aspect angle.
In other words, impedance loading is a technique which introduces an echo source to
cancel the back-scattered field. However, the required impedance value is highly
dependent on the structure of the target, location and distribution of the loads, operat­
ing frequency, and observation aspect. Therefore, fixed linear impedance loading is
not effective at reducing the RCS over broad spectral windows and wide aspects, and
this technique has been discarded in RCS reduction application.
Nevertheless,’ impedance loading can cause other interesting effects, for exam­
ple, it might change the natural frequency of a target [84] or alter the range profile
and image’s appearance. Recent advances in technology make it possible to control
the load electronically. This time varying loading can introduce other phenomena.
Examples are: making the receiver in a coherent radar unable to phase lock to the
frequency of the incident wave, shifting the apparent frequency of the scattered field
to provide a false Doppler shift [85], and spreading the spectrum of the scattered
field to decrease energy within the bandwidth of the receiver.
This latter
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 174-
phenomenon can be viewed as thermalization of the incident wavefield. The degree
that the scattered field is changed by introducing modifiable loads will clearly deter­
mine how pronounced the impedance loading affects the above phenomena.
A load can either be linear or nonlinear and is characterized by the voltagecurrent curve (v - / curve) across its terminals. Linearly and nonlinearly loaded Nport scatterers have been analyzed [86,87,88,89,90,91]. Numerical techniques are
usually required to theoretically calculate the field scattered from a loaded scatterer of
arbitrary shape; the moment method is the most popular of these. It is well known
that the computing time of the moment method is exponentially proportional to the
dimension of the object in terms of the operating wavelength. For microwave diver­
sity imaging, the object is usually large compared to a wavelength and multiple fre­
quencies have to be used. This is equivalent to adding another dimension to the
problem. It is therefore impractical to numerically analyze the effect of loading on
the scattered field and image of a 2-d or 3-d object using frequency diversity imag­
ing. If the object is confined to a thin rod, however, the computing time will be
tolerable if the numerical analysis program is sufficiently sophisticated. Furthermore,
edges of a complex object may be approximated by thin wires in certain situations
and will be further discussed later in this chapter. Besides, when a thin rod is
illuminated near.^he end-on direction, the scattered field is contributed by a special
scattering mechanism, a surface traveling wave [21]. It would be interesting to
examine the image created by this special scattering mechanism. For the above rea­
sons, a thin rod will be the primary object to be numerically and experimentally stu­
died in this chapter and will serve as a vehicle for gaining insight in the difficult
problem of nonlinear loading of a complex body for the purpose of modifying its
RCS. Monochromatic imaging of a monopole antenna has been studied holographi­
cally by Iizuka [14], who was interested in visualizing resonance effects. However,
we are interested in a wire scatterer rather than an antenna and frequency diversity
imaging instead of a monochromatic holographic image.
In this chapter we will review the scattering theory of scatterers with linear
loading, nonlinear loading, and time-varying loading.. A thin rod will be used as the
test object. Numerical and experimental results pertaining to the scattered field and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 175 -
images of a thin rod will be demonstrated and discussed.
7.1. Linear Impedance Loading
The effect of linear impedance loading on the back-scattering of a conducting
object has been extensively studied. Chen and Liepa [92] presented a theoretical and
experimental study on the minimization of back-scattering of a thin cylinder with
central loading. Yu and Shen [93] used multiple impedance loading to modify the
scattered fields of a thin wire. Exact expressions of current distribution on the wire
were formulated. For targets of other shapes, the field scattered from a conducting
object with impedance loading can be expressed in terms of a load-independent quan­
tity and a load-dependent quantity by utilizing N-port network theory. Harrington et
al. analyzed the scattering from loaded N-port scatterers [86] and determined the
characteristic modes of a loaded scatterer [90]. They applied the concept of charac­
teristic modes of a loaded body to control of radar scattering by reactive loading
[87,88]. They also gave a procedure for obtaining the reactive loads of an N-port
loaded scatterer which maximize the radar cross section [89].
Although the back-scattered fields could be maximized or minimized by lumped
impedance loading, the load values are highly dependent on the frequency, loading
positions, number of loading points, and aspect or viewing angle. Loading can be
lumped or distributed. Wu and King [94] reported that an antenna can be made
reflectionless if it is resistively loaded such that its internal impedance continuously
changes from the center-fed point to the ends and satisfies a certain form. Their
theoretical results .were verified experimentally by Shen [95], who approximated the
continuous resistive loading by a linear stepped-function approximation. Rao et al.
[96] demonstrated experimental results of the broadband characteristics of a cylindri­
cal antenna with exponentially tapered capacitive loading. However, the problem we
are dealing with involves a scatterer rather than an antenna, with a lumped loading
instead of distributed loading.
In the following, an N-port network analogy of a loaded N-port scatterer will be
summarized.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 176 -
7.1.1 Analysis of a Loaded N-port Scatterer
A loaded scatterer is one which has one or more ports terminated by admittance
elements. An N-port scattering system can be represented by a general network
diagram as shown in Fig. 7.1. Denote the mutual admittance between port k and port
j at angular frequency CD as Ykj (to), which is defined as the short circuit current at
port k divided by the voltage at port j with all other ports being short circuited, that
is,
>
i
N-port Linear
Netwoik System
b
Fig.7.1 Equivalent network of an N-port loaded scatterer.
=
/*(«» i
k (« ) = °
for all i * j
(7.1)
If the voltage at each port is Vj((o), then by superposition, the current at port k will
be
h «Q) = " £ Ykj (co)V, (co) - lgkE(co)Einc (co)
(7.2)
where Igkg is the short circuit current at port k due to an applied unit incident field
Einc (co) with all ports being short circuited. The scattered fields of the loaded
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
scatterer can be written as
N
Es (co) = X Egj (CO)Vj (co) + EgE(co)
>=l
(7.3)
where Egj (co) is the scattered field due to a unit voltage applied at port j with all
other ports being short circuited, Vj (co) is the voltage across port j , and EgE (co) is
the scattered field of the scatterer without loading. For linear impedance loading, the
port current and port voltage are related by
4 (co) = y*(co)V*(co)
(7.4)
where y*(co) is the load admittance at port k. Substituting Eq.(7.4) into Eq.(7.2) we
have the following matrix relation
(7.5)
where [ Y ] is a matrix with elements l*; (co), [ y ] is a matrix with elements
ym(co)
m=n
(7.6)
m*n
and [ lE ] is a vector matrix with element
(7.7)
ImFM = -/*«£(©)•£,« (®)
From Eq.(7.5), the voltage vector [V] can be obtained from the relation,
[V] = {[K] + Lv]}-1 [IE]
(7.8)
Substituting Eq.(7.8) into Eq.(7.3), the scattered field becomes
Es(co) = [Eg]T-{[Y] + [y])"1^ ] + EgEm
(7.9)
where [ Eg ]T is the transpose of the vector [ Eg ] with element Egl (co). It is noted
that the quantities [F], [Eg], [IE], and ^ ( c o ) are independent of the loaded
impedance, the only load-dependent quantity is the matrix [y].
If only one loading point is present, the scattered fields can be simplified as
(7.10)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 178where F (co) is the input admittance of the scatterer at the loading port, and y/,(co) is
the loaded admittance. Eq.(7.10) can be abbreviated in the following form
E.(co) =X(a>)A (to) + B (co)
(7.11)
= vy (f»)
7 T+ —
TT
yL (to)
<7-12)
A (co) = IgE (co)Einc (co)Eg (co)
(7.13)
with
fi(co) = £g£(to)
£ ,( co) can be nuUified if the real part of the solution of yE(co) in Eq.(7.10) is nonnegadve. If the real part of the solution is negative, then an active load is required to
nullify the scattered field. Assume the load is constrained to be passive, £,(co) can
have both maximum and minimum values if yL (co) values are properly chosen. Green
expressed the scattered fields of a scatterer loaded with ZL at one-point in a different
form [97]
Es (Zl ) = Es (Z*) + r;/(Z * )£ r
(7.14)
=ES(Z*) + r v V(Z*)Er'
(7.15)
where ES(Z*) is the field scattered by the object with a conjugate matched load, E r
is the field scattered by the object if excited by a unit current source at the load
point, Z* is the conjugate of the equivalent impedance seen when looking into the
load terminals, I(Z*) is the current flowing through a conjugate matched impedance
at the load terminal when the object is in the presence of the incident field, V(Z*) is
the voltage across the port terminated with conjugate matched impedance, E r is the
field scattered by the object when excited by a unit voltage source at the load point,
and Tj and IV are the reflection coefficients at the loading point and are given by
r*
r, =
ZL ~ Za
~~— ~ ~
-a + Zt
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(7 .1 6 )
- 179 and
rv -
C7.it >
Z*(Za +ZL)
It is seen that Ej(Za),/(Z a), Er , V(Za), and E /
are independent of the load
impedance. The only load-dependent terms are T/ and Fv . For a passive load the
absolute value of the reflection coefficient is less than or equal to 1. If
I ES(Z*) I 5 V{Z*)E/ ,
then the scattered field can be nullified by a load with reflection coefficient
ES(Z*)
-
r v = (7-18)
V(Z*)E/
If
IE*(Za*)l > W i Z ^ E ' I
tlicn
l£A.I can not be nullified.However,
minimized. In die following, we
the Mattered field can bemaximized or
willshow
that theextremevalues
are on the
ITI = 1 circle, which implies that the load is purely reactive.
Let
ES(Z*) = a x + j b x
V(Z*)Er' = a 2 + j b 2
r v = * +Jy
where a x, b i, a 2, b 2, x , and y are all real. We then have the following problem:
Find x and y , such that
\Es (Zl )\ = f { T v ) = f ( x , y ) =
I (a x+jb {)+(x+jy )(a2+jb2) I
(7.19)
has extreme values subject to x 2 + y 2<\.
The solutions are as follows:
1.
If a y + b y < a 2 + b 2 , the minimum value of f ( x , y ) is zero, which occurs when
~(aya2+byb2)
X° = --------2TT2—
’
aj+ bj
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(7,20)
- 180 and
aib2-a2bi
y° =
2.
a iT+TbTj T
(7-21)
If a 2+b2>a2 +b2 , the extreme values occur when the following boundary con­
ditions hold
i n = ljt2+y2 l 1/2= 1.
(7.22)
In this case, the load is purely reactive. Let y 2 = I - * 2, then
f 2 = \ E s \2
= (a 2+b 2 )+(a 2 +b2 ) + 2(a ta 2+b \b 2>x+2 (ai~a i&^ l - x 2
(7.23)
Take the partial derivative of/ 2 with respect to x» and set it to zero, we then
have
d f2 „
mi i_ \
— = 2(0^2+^ 1^ 2) ----------- 1
ax
\
n
” ®
(7.24)
(g 1a 2+ b 1b 2)2
!)(a|+b2Z)
i ^ ( 2 L ) ' m a x = m a x ( / ( * o ) » / ( “ ■*()))
( 7 -2 6 )
= min( f ( x 0) , f ( - x 0)}
(7.27)
From the solutions of x 0 and y 0, we can find the required ZL from Eq.(7.17).
It is noted that the solution of ZL which minimizes or maximizes the scattered
field depends on the locations of the loading points. Es is a function of N complex
variables if the number of loading points is N. There are N degrees of freedom to
modify the value of Es and the solutions of the null fields or of the extremes may
not be unique. If all the impedance values of each loading point are restricted to be
the same, Eq.(7.9) is then a function of one complex variable. For that case we can
find the extreme values of the scattered fields of the loaded scatterer.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 181 -
To obtain the scattered fields of the loaded scatterer, the quantities
[Eg], [T], [IE], and EgE in Eq.(7.9) must be determined. All those quantities are
dependent on the geometry of the body. Exact expressions of these quantities can be
obtained only for certain simple objects. Numerical techniques, however, can be
applied to a body of arbitrary shape and are generally only limited as stated earlier
by the wavelength size of the body. The moment method is one of the most popular
numerical techniques to solve this EM problem. In this section we will numerically
investigate the scattering properties of a linearly loaded scatterer. A thin straight wire
is the simplest geometry and will be used as an example. A review of the formula­
tion for a straight wire with impedance loading follows.
7.1.2 Formulation of a Straight Wire with Impedance Loading
Consider a cylindrical wire with radius a and extension from z= -/ to z=l along
the z-axis as shown in Fig.7.2. Assume the radius is small compared to a wavelength
so that the only significant component of current on the wire is the axial component.
Denoting the current on the wire as I ( / ) , the z-component scattered field due to this
current distribution can be written as [44]:
El = —
f[-^ (Z-’Z ) + * V (z.z')]/0 O dz'
ycoeo -/
dz2
(7.28)
where V|/(z ,z ) is the free space Green’s function,
with k the wavenumber in free space; R is the distance between the observation point
(x ,y ,z) and the source point (x’ ,yf ,z'), and is given by
R = ^ ( x - x ' ) 2+ ( y - y ) 2+(z-z' )2
(7.30)
If the incident field is Zf , the boundary conditions require that the tangential com­
ponent of the total field, (i.e., the sum of the scattered field and the incident field)
along the wire surface be zero (£/(z) = - E ‘(z)), or
f/(z')
ytoeo-/
dz2
+ kMz,z')]dz' = - £ '( z )
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(7.31)
- 182z
i>
A
1
z
.
zn
zb
^
1
"
T
'
'
I
=
1
o
II
J Z = -l
-H I*
a
Fig.7.2 Geometry of a straight wire.
The above equation is called Pocklington’s Equation [44]. Ez‘(z) is usually known or
can be evaluated. The only unknown parameter to be obtained is the current distribu­
tion 7(z'). For convenience, we rewrite Eq.(7.31) in the following form
z
\l(z)K (z ,z') dz ' = - E zl (z)
-z
(7.32)
Assume that the current I (z ) is approximated by a series of expansion functions Fn
such that
I (z ') = h n W )
(7.33)
n -1
where /„ is a complex expansion coefficient, and Fn(z') is an expansionfunction
which is nonzero only for z' in Az„. SubstitutingEq.(7.33) intoEq.(7.32), and letting
z =zm, one obtains
£ / „ j Fn (z' )K (zm,z' )dz = -Ei(zm)
n=1
Az,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(7.34)
- 183 -
where the subscript m on zm indicates that the integral equation is being enforced at
segment m. Expressing the integral part of Eq.(7.34) as Z ^ , and the right part as
Vm, one has the following compact form
= Vm
m = 1,2
(7.35)
M
n=l
It is noted that the EM integral equation has been reduced to a set of linear simul­
taneous equations. Expressing Eq.(7.35) in matrix form we have
[zm
] [ ',]
= [^]
(7.36)
The matrices f z ^ 1, j/„ j, and jym j are refened to as the generalized impedance,
current, and voltage matrix respectively. However, the units of Z ^ ,, /„, and Vn need
not necessarily be ohms, amperes, and volts respectively.
The boundary conditions can be enforced at a point on each segment or in a
weighting average sense over each segment. Denote Wm(z) the m ,h weighting func­
tion. The boundary condition becomes
jWm (z )[E/(z) + E ‘(z )]dz = 0 m =1 ,...A
(7.37)
By substituting Eq. (7.34) into Eq. (7.37), one has
N
1
1
Z A ,J w«(z) i W ) K ( 2 ,z')dz' dz = - j W m(z)E‘(z) dz
71=1
-/
Azm
(7.38)
-I
which can be abbreviated by
= Vm
(7-39)
71=1
If the weighting function and the expansion function a.e in the same forms, the pro­
cedure is called the Galerkin method [35]. It is noted that integration over
z
results
in the elements of the generalized impedance matrix and generalized voltage matrix
having volts as their units. It is noted that Z ^ is independent of the incident waves;
it is only a function of the geometric structure of the scatterer or antenna, expansion
function, and weighting function being used. The Choice of a desired expansion
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 184-
function and testing function has been widely discussed[36]. The most popular ones
are pulse expansion with impulse weighting, 3-term trigonometric expansion with
impulse weighting, overlapping triangle expansion and weighting, and overlapping
piecewise sinusoidals [37].
The physical meaning of the element Zmn can be interpreted as "the weighted
scattered fields integrated over the segment Azm due to a source at segment Az„ with
weighting function Wm(z) and source expansion function F„(z')." Once [Z ^ ] and
[Vm] are known, the current distribution along the wire can be calculated by
[/] = [zm j '1 [ v . ] = [rm ] [v „ ]
where
(7.40)
j = jZfm J 1 and is called the generalized admittance matrix. The physi­
cal meaning of the elements of Jy^, j is the mutual admittance between port i and
port j defined as the short circuit current at port i due to a unit voltage across the port
j when all other ports are short circuits.
For the specific geometry of a straight wire, the generalized impedance matrix
has a special property. All the values of the N 2 matrix elements are contained in any
one row of jz^ , j, say the first row, if all segments Azm are of equal length. All
other rows are merely a rearranged version of the first row. The elements satisfy the
following relationship
Zm n= Z \,\m-n i+l
n2 1
(7.41)
Such a matrix is called a toeplitz matrix. Efficient algorithms are available to calcu­
late the inverse of a toeplitz matrix. The computing time and storage required for
solving a toeplitz matrix is much less than that for solving a non-toeplitz matrix [44].
The inverse of a toeplitz matrix, however, is not in toeplitz form [85].
A scatterer or an antenna with N segments can be viewed as an N-l port net­
work. At each port, a generator and/or impedance (active or passive) can be con­
nected in series or in parallel to the port. The N-segment wire and its equivalent circuit are shown in Fig.7.3. If a load Zm and a generator V* is inserted into the m
junction having a current lm, the total voltage at that port is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
th
- 185-
L
EC
EC
EC
port
port 2
port 3
m th port
port 4
Fig.7.3 N segment loadedwire (a) N-l port tenninals pairs, (b). Equivalent circuit for
the m th port.
Vv m = Vs
y m - i1m Z
Cjm
(7.42)
The m th equation in a system of N linear equations will become
iz -./.-v a n-
Imzm
(7.43)
1
or
2
(7.44)
= vi
n- 1
where
Zm-,,
^nvn
(7.45)
= Zmm
nun “I" Zm
m
Z__'
nvi —Z__
tnn
for m&n
(7.46)
It is seen that the effect of lumped loading is simply that of adding a load impedance
Zm to the corresponding diagonal elements in the impedance matrix. The impedance
matrix, however, is no longer toeplitz, and the efficient program for solving a toeplitz
matrix cannot be applied directly. Usually, the number of loading points is much less
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 186 -
than the number of segments. The new solution for the current vector can be
expressed in terms of the current solution without loading and the loaded impedance
[85].
Suppose there are K loading points with impedance
serially connected to the
m ‘ih port. Denote Z the generalized impedance matrix without loading, Z ' the matrix
with element
Z
zm' form e {m,}
' = ' 0 otherwise
Zmn ~ 0
(7-47)
f°r m * n
(7.48)
We have
(Z + Z ')7 = V
(7.49)
Let 7° be the current solution vector without loading, that is,
Z7° = V
*
(7.50)
7° = Z~lV =YV
(7.51)
where7° is the vectorof the current solution without loading. Multiplying Eq.(7.49)
by T, we have
Y(Z+Z')I = YV
(7.52)
(U+YZ')I
(7.53)
=YV
where U is the N.xN unit matrix. After expansion of Eq.(7.53), we have the follow­
ing linear equation
( l + n , ^ ' wwV«, + £ r « , ^ . / 1^ = / »
i*i
In = 7„ ° - £ YnmZ m.m.lmi
/=
n not in {mxrn%..., mK}
(7.54)
(7.55)
i= 1
It isnoted that the number of unknowns in the linearly simultaneousequations of
Eq.(7.54) is K (the number of loading points) instead of N (the total number of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 187-
segments). The computing time required for solving K unknowns is much less than
that for solving N unknowns. After
are solved, the current solution
/„ ( n not in {m, }) can be expressed in terms of the {/„.} and elements of [ 7° ],
the current solution vector without loading.
7.1.3 Scattered Field and Images of a Thin Rod Scatterer
In the following, we will use the moment method to find the effect of
impedance loading on the scattered fields of a straight wire scatterer and its recon­
structed images. The piecewise sinusoidal Galerkin method is used. The piecewise
sinusoidal function is defined as
sin[A: (z — x)]
s in tk ^ -z ,,.!)]
sin[k(zw+1-z)]
*«<*> = sin[fc (zn+1-z „ )]
zn - 1
(7.56)
zn<z<zn+1
(7.57)
and is shown in Fig. 7.4. The elements of the generalized impedance matrix using
the above expression and weighting function can be shown to be [44]
^mn ~
*: SW i z - z ^ Q ] + Y sin[k(zm+1-z)]
'-1 sin(fcAzm)
* sin(&Azm)
Zm
(7.58)
-jkR.
-jk R .
Rn - 1
j 30
sin(fcAz_)
- 2cos(kAzn)Rr
R n+1
dz
where
R n -1 =
^Ja2 + (z-zn_i ) 2,
Rn =
R n -1 =
^ a 2 + ( z - z n )2
^la 2
+
(7.59)
(z-zn+1)2
= zn ~ zn -1 = zn+l ~ z n » ^ zm ~ zm ~ zm -l = zm+l — zm
(7.60)
When the angle of incidence is 0, the elements of the generalized voltage vector
become
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 188-
Fig.7.4 (a) Piecewise sinusoidal expansion function, (b) Set of overlapping piecewise
sinusoidal expansion functions.
Vm =
(e'/tem‘lCOS0 —eji*mGOS0[cos(/: Azm) - j cos0sin(fc Azm)]
s\n(k Azm)k (l-cos20)
+ ejh m+1cose
_
e jkxmcosejcos^ Azm) + y cos0sin(itAzm)]}
(7.61)
The impedance dependence of the field scattered from a loaded straight wire
will be demonstrated. The maximum and minimum scattered fields can be obtained
by suitably choosing an impedance value for a given number of loading point and
loading position as previously shown. Noted in Fig. 7.5 are the maximum and
minimum values of the back-scattered field, which are obtained by loading a suitable
passive impedance at the center point, versus the rod length in terms of incident
wavelength at several angles of incidence. The scattered fields without loading (i.e.,
short circuits) are also shown in the figures. In the broadside (see Fig. 7.5(a)), the
difference between Esmax (co) and Esmin (co) is not very great when the wire length is
greater than 1.4 wavelengths. Apart from the broadside, scattered fields can be
nullified for some frequencies at some aspect angles. This can be explained from Eq.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 189-
(7.10). If the second term EgE (co) of Eq. (7.10) is much greater than the maximum
value of the first term, the effect of loading on the scattered field will be very small.
Apart from the broadside, i.e., in the side-lobe region, EgE(co) can be very small.
Therefore, it is more possible to nullify the back-scattered field in the side-lobe
region than in the broadside. The required impedances (real part and imaginary part)
to nullify the back-scattered fields for angle of incidence equal to 9 = 45° are shown
in Fig. 7.6. If the real part of the required impedance is negative, an active load is
required to nullify the field. Shown in Fig. 7.7 are the passive impedances required to
maximize and minimize the back-scattered field at 0 = 45°. As analyzed in Sec.
7.1.1, the impedance required to optimize the field is purely reactive if the backscattered field cannot be nullified. Observing the plots of the above figures, one
notes that the required impedance is highly dependent on frequency. Plots of the
required impedance at other 6 (not shown here) also shows the same frequency
dependence. From the above figures, one can conclude that fixed impedance loading
is effective at reducing or enchanting the RCS only at certain frequencies and aspect
angles.
As shown in Fig. 7.5, the back-scattered fields can be reduced a great deal
through a central passive loading for most frequencies when the aspect angle differs
from broadside. At broadside, the difference between the maximum value and the
minimum value of the scattered field is small (see Fig. 7.5(a)). Now, we will increase
the number of loading points and study their effect on the broadside back-scattered
fields. We assume that each loading impedance is of the same value and is reactively
varied between -j200Q and y'200Q in 10Q steps. The maximum, minimum, and noload values of the broadside back-scattered field are shown in Fig. 7.8 for the number
of loading points nL equal to 3, 5, and 7. The positions of the loading points are
also shown in the accompanying figures. From the figures shown, one can note that
additional equi-spaced and equi-amplitude loads are fairly ineffective at changing the
broadside RCS.
Next, we will examine the effect of impedance loading on the range profiles and
images. Range profiles can give useful insight into the scattering mechanism. Radia­
tion can originate from several places on an arbitrarily shaped wire object. These
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 190 2.0
o
-3
3
C.
0 .5 r" \ .
2 .3
1.1
3 .5
4 .8
6.0
length in terms of wavelength
(a)
.3 2
.2 4
•8
.2L
i.
E
*
.0 8
3 .5
2 .3
1.1
4 .8
6.0
length in terms of wavelength
(b)
05
amplitude
^m ax
.0 3 7 5
E (Z
/
.0 2 5
.0 1 2 5
0
1.1
^ m in
1
' v •'
.S ’
/
------------- .S C ./- .......
\ A VA -'
_____
3.5
/
7
V
y
2 .3
= 0)
l
0
/
/
4 .8
AV
Aw.
6 .0
'
length in terms of wavelength
CO
F ig. 7 .5 (cap tion se e n ex t p a g e)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 191 -
E(Zl
= 0)
max
mm
a
4.8
2.3
3.5
length in terms of wavelength
6.0
(d)
2.4
1.2
0.6
1.1
4.8
2 .3
3.5
length in terms of wavelength
6.0
(f)
1.2
=0)
^mix
A
E{Zl
4>
■aa 0.9
|
"^ m infi\
0.6
0.3
l.l
4.8
3.5
2.3
length in terms of wavelength
(e)
6.0
F ig.7.5 Extreme values and no-load values o f the back-scattered field o f a straight wire
scattered versus length in w avelength at angle o f incidence equal to (a) 9 0 °, (b) 75°,
(c) 60°. (d) 45°. (e) 30°. (f) 15°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 192 -
real part
imaginary part
V)
£
JZ
o
oo
a
ea
T3
&
2.3
3.5
4.8
length in terms of wavelength
Fig.7.6 Impedance required to nullify the back-scattered fields for angle of incidence
equal to 0=45°.
400
•3
.S
200
i -400
1.1
2.3
3.5
4.8
6.0
length in terms of wavelength
Fig.7.7 Required passive impedance to maximize and to minimize the back-scattered
fields for angle of incidence equal to 9=45°.
Real part of ZL(Emin)
Imaginary part of ZL (E ^
Imaginary part of ZL(Emax)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 193 -
Cl
0.5
1.1
4.8
2.3
3.5
length in terms of wavelength
(a)
6.0
2.0
o
1.0
2
1*
s
CQ
1-0
0.5
1.1
2.3
3.5
4.8
6.0
length in terms of wavelength
(b)
0
E{Z, = 0)
2.3
3.5
length in terms of wavelength
Cc)
F ig .7 .8 E x trem e v a lu e s an d n o-load v a lu e s o f the b road sid e back -scattered fie ld s o f a
straight w ire v ersu s len gth in term s w a v elen g th w h en the num ber o f lo a d in g p o in ts is
(a). nL= 3, (Z[ = 0 .5 / = - z 3 z 2= 0 )
(b). nL= 5 . (Z] = 0 .6 7 / = -Z j, z 2 = 0 .3 3 / = - z 4_ z 3 = 0 ).
(c). nL=7, ( z [ = 0 .9 5 / = - z 7.
= 0 .6 7 / = - z 6 z 3 = 0 .3 3 / = - z 5 z 4 = 0 ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 194 -
include the excitation region, an impedance load, a change in radius, a sharp bend, a
smooth curve and an open end [38]. Consider a straight wire illuminated by an
impulsive plane wave with angle of incidence 0 as shown in Fig. 7.9. In this scatter­
ing arrangement the only places which cause radiation are the end points of the
wire. The pulse traveling in free space will impinge on the upper end point first. Part
of the incident energy will be re-radiated and the remaining energy will continue to
travel along the wire. This traveling pulse will be partly re-radiated when it reaches
the lower end and partly reflected upward along the wire. This process of radiation
and reflection will continue until the pulse dies out. The original pulse propagating in
free space will hit the lower end point some time after it impinges on the upper end
point. The process of radiation, reflection, and guiding propagation along the wire
will then occur just as in the case of the upper end point. The differential path
lengths of radiation occurring at each instant relative to the path length when the
impulsive illumination hits the center point of the wire are as follows:
at 1
I j = - h cos0 - h cos0 = - 2 h cos0
at 1' l { = hcosB + h cos0 = 2/icos0
at 2
l 2 = -hcosQ + 2 h + /icos9 = 2h
at 2'
12 = /icos9 + 2h - /icos9 = 2h
at 3
/ 3 = -/tcos0 + 4h - /tcos9 = 4h - 2/tcos9
at 3'
/ 3' = heos0 + Ah +hcosd = 4h + 2Acos0
at 4
I 4 = -/icos9 + 6 h + /icos9 = 6 h
at 4'
1/
= ftcos0 + 6 h - hcosQ = 6 h
Let the length of the wire be 30 cm, and the frequency coverage be from 6GHz
to 16 GHz. In other words, the length in terms of wavelength is from 6 to 16. The
polarization of the incident field is assumed to be 0-polarized. Shown in Fig. 7.10 are
the magnitude of the range profiles for angle of incidence equal to 30°, 45°, 60°, and
75°. If we carefully examine the range profiles shown in Fig. 7.10, we can find that
the peaks marked with 1 and 1' depart more from the center as the angle 0 decreases,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 195 -
incident
illumination
illumination
z —-h
z = -h
Fig.7.9 A straight wire illuminated by an impulse.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 196 -
while the peaks marked with 2 and 2', and 4 and 4' remain at the same position and
thus are independent of the angle of incidence. Those peaks marked with 3 move
toward 2, while those marked with 3' move toward 4' as 0 increases. These obser­
vations verify the above analysis. It is interesting to note that the ratio of the ampli­
tude of peak 1 to that of peak 1' decreases as 6 decreases.
A real thin rod with length 12" and diameter 1/8" is used as a test object to
experimentally verify the above numerical results. The frequency coverage is from 6
GHz to 16.5 GHz and the waves used are co-polarized. It’s range profiles at several
aspects are shown in Fig. 7.11. It is noted that a fixed bistatic angle about 16° exists
between the transmitting antenna and receiving antenna in the experiment. In this
bistatic case the differential path length of path 2 and path 2’ will not be equal. This
fact explains the discrepancy between the experimental and numerical range profiles.
From the previous analysis and the range profiles shown in Fig. 7.10, one can
see that the phenomenon of traveling waves is quite evident in the straight wire case.
If the rotation center is chosen at the center point the effect of the constant ranges
(2 h , 6 h , etc.) on the reconstructed image will be a ring with constant radius as
explained in Chap. 3. Shown in Figs. 7.12(a) and (b) are the numerical and experi­
mental fringe patterns of the real part and imaginary part of the scattered fields
respectively, while Figs. 7.12(c) and (d) are the numerical and experimental images
reconstructed over an angular window with 0 from 20° to 80° respectively. It is seen
that a ring appears in the images and the end points are intensified. This example
shows that the presence of traveling waves usually degrades the image.
If
three
lumped resistors,
each
with
resistance
50£2,
are
added
at
z {=0.5l, z2=0, z 3= - 0 .5/, these loading points will cause extra reflections. Both the
incident wave impinging on the loading points and the waves traveling along the wire
arriving at the loading point will cause additional reflections. This fact results in
additional peaks in the range profiles. Shown in Fig. 7.13 are the range profiles of
the 3-loading-point wire at several angles of incidence. Examining these plots, one
finds that more lobes appear and the lobe produced by the loading point is not as nar­
row as those produced by the end points. Furthermore, the number of lobes between
1 and I ' is not necessarily equal to the number of loading points (for example, see
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 197 -
0.075
S 0.10
0
37
-37
distance in cm
distance in cm
(b)
(a)
0.1
•u
.
0.1
0.075
. . .
-
1"
u 0.075 •
■a
/
a
U 005
St 0 05
•
$
03
a
0.025
§
0.025
.
-37
-37
distance in cm
(c)
—
-r
i
0
II
J l
J
3
'
A*
'
37
74
distance in cm
(d)
F ig .7 .!0 T h e nu m erical ran g e profiles o f a straight w ire for angle o f incid en ce equal to
(a) 30°, (b) 45 °, (c) 6 0 °. an d (e) 75°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
F ig .7 .11 T h e ex p erim en ta l ra n g e p ro files o f a straight w ire fo r angle o f in cid en ce equal
to (a) 3 0 ° , (b ) 4 5 ° , (c) 6 0 ° , and (d) 7 5 ° .
-199 -
d i p . 0 ; < 0 .8 0 )d e g .
max : 2 . 0 ; d l * p ( . 5 , - . 0 6 )
a r^
n
b
p o d . 0 ; (0 / 9 0 ) d e g
max; 7 . 2 ; d l o p ( 2 . - 2 ) "
r**l podi (it.T fttfe g
w < E E i dl*p (66.30)
d i p . 0 ; <0,00)deg
m a x ;6 .l;d i« p (,5 .-.0 6 )
r o d , 0 ; <0.90)deg
max: 1 4 ; d i s p ( 2 , - . 2 )
tlaiUtad P6di (le.TVldM
«*x>45j ditp (4s.a»>
F ig .7 .12 N u m e ric a l a n d e x p e rim e n ta l frin g e p a ttern s an d im a g e s o f a th in ro d . (a)
R e a l p a rt (left) a n d im a g in a ry p a rt (rig h t) o f th e n u m e ric a lly b a c k -sc a tte re d field p a t­
te rn . (b ) R e al p a rt (le ft) an d im a g in a ry p a rt (rig h t) o f th e m e a su re d b a c k -sc a tte re d
fie ld p a ttern , (c). N u m e ric a l im a g e re c o n stru c te d o v e an a n g u la r w in d o w fro m 0 = 2 0 °
to 8 0 °. (d). E x p e rim e n ta l im a g e re c o n stru c te d o v e an a n g u la r w in d o w fro m 0 = 2 0 ° to
8 0 °.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 200-
Fig. 7.13(c), with 0=45°). These additional lobes are due to reflections of the travel­
ing waves.
To experimentally examine the loading effect, we divide the thin rod mentioned
into three sections with a 1 mm gap between sections. These gaps are expected to
produce a loading effect. However, it is difficult to assign a loading value in each
gap. Additionally, the equivalent loading impedance is also a function of frequency
because the gap distance in terms of wavelength is changed with frequency. The
experimental range profiles for several angles of incidence are shown in Fig. 7.14.
More peaks appear in the range profiles due to the additional discontinuities in the
gaps. However, the magnitude of these peaks differs from the counterparts of Fig.
7.13.
The numerical and experimental fringe patterns of the real part and imaginary
part of the scattered fields from the above loaded scatterers are shown in Fig. 7.15(a)
and (b) respectively. Their numerical and experimental images reconstructed from an
angular window from 0=20° to 80° are shown in Fig. (c) and (d) respectively. It is
seen that the loading impedance and the surface traveling waves have distorted the
images. By comparing the images of Fig. 7.15 and Fig. 7.13, one can conclude that
the images have been successfully distorted by impedance loading. However, the
price paid is an increase in the RCS. The measured mean RCS patterns with and
without gaps averaged over the whole bandwidth are shown in Fig. 7.16.
7.2 Nonlinear Impedance Loading
The property of a load can be characterized by the voltage-current curve (v-t
curve) across a load. For a linear load, the v - i curve is a straight line. If the v - i is
not a straight line, then the load is called a nonlinear load. Typical examples of non­
linear ioads are diodes for switching, thresholding, modulation, and harmonic genera­
tion.
Nonlinear effects are important in the area of electromagnetics. For example:
nonlinear effects have to be considered when an antenna system containing semicon­
ductor devices is illuminated by a strong lightening stroke or a nuclear
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 201 -
.1 0
<i>
3
T3
i
jiA a
.025
/I L —
0
-37
37
(a)
(b)
74
.10
-
•
U .075
•a
2
|
37
distance in cm
.10
*, -075
■a
0
distance in cm
3
.050
U -050
.025
.025
-37
0
37
distance in cm
(c)
74
-37
0
37
distance in cm
(d)
F ig .7 .13 N u m erical range profiles o f a straig h t w ire w ith three load in g p oints at angle
o f incidence e q u a l to ( a ) 30° (b) 45°, (c) 60°, and (d) 7 5 ”.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
-202
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-
F ig .7 .14 E x p erim en ta l range p r o file s o f a straigh t w ir e w ith th ree lo a d in g p o in ts at
a n g le o f in c id e n c e eq u a l to (a) 3 0 ° , (b ) 4 5 ° , (c )
6 0 ° , and (d ) 7 5 ° .
- 203 -
d i p . 3 ; (0 ,7S )deg
d i p . 3 ; ( 0 ,7 S ) d e g
max ■1 , 8 ; d l s p ( . 5 , - . 0 6 ) max :6.2;dlsp(^S,__-p05)
a
r o d . 3 ; ( 0 .9 0 ) d e g
m ax ;4 .3 ;d lsp (2 .-,2 )
r o d . 3 ; C0,90)deg
max: 1 4 ; d l s p ( 2 . - , 2 )
b
r**l rodj 4 tegM niftj tl0*70)d»g
mx
>GB; d t* p (58*33)
d
ilMul*t*d rod) fiL*3i ft»60 ohm
mx
>48; d l* p (48*33)
c
Fig.7.15
Numerical and experimental fringe patterns and images of a thin rod with three
loading points, (a). Real part (left) and imaginary part (right) of the numerically
back-scattered field pattern, (b). Real part (left) and imaginary part (right) of the
measured back-scattered field pattern, (c). Numerical image reconstructed ove an
angular window from 8=20° to 80°. (d). Experimental image reconstructed ove
an angular window from 9=20° to 80°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 204 -
3
M
P
1
%
U
*
e
i
n
d
b
hV
E .3 :
RGD0 AND R0D3
Fig.7.16 Comparison of the measured mean RCS patterns of the thin rod with and
without gaps.
electromagnetic pulse [90]; a scatterer with nonlinear loading may spread the spectra
of the scattered field into several harmonics, "thermalizing" thereby the scattered
field..
Nonlinear effects for a wire scatterer and a dipole antenna have been extensively
studied. Sarkar and Weiner [91] applied the Volterra series technique to analyze a
nonlinearly loaded antenna. The advantage of that method is that individual fre­
quency components in the output can be determined directly without performing a
fast Fourier transform on the total time domain. The Volterra series technique is use­
ful if the nonlinearity is not too strong. Schuman [98] used a space-time-domain
integral equation to treat a nonlinearly loaded antenna. Liu and Tesche [90] used two
methods for analyzing antennas and scatterers having nonlinear resistive loads. The
first approach involves the use of a direct time-domain integral equation. The second
uses frequency domain data to compute the time-dependent currents and voltages
across the nonlinear loads. The nonlinear problems are then solved by time-stepping
and convolution utilizing the solution of the linear portion of the network. Djordjevic
and Sarkar [99] used a method similar to the method of [90] but with a different
approach, which requires less time-domain data or a shorter duration of Green’s func­
tion. Laudt and Miller [85] used the time-domain integral equation to study the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 205-
behavior of a wire antenna or wire scatterer loaded with nonlinear elements, the non­
linear load types considered include those with piecewise-linear voltage-current
curves and a load with time varying resistance.
In this section we wili first review the analysis of an N-port nonlinearly loaded
scatterer and then numerically discuss the scattering properties of a nonlinearly
loaded thin rod scatterer.
7.2.1 Analysis of an N-port Nonlinear Loaded Scatterer
Consider an N-port network. Assume an ideal voltage source v; (r) is applied
across the port j , and all the other ports are short circuited. The short circuit current
at port k due to vy-(r) can be expressed as
Ik (co) = Ykj (co)Vj (co)
where
V j ( ( 0)
(7.62)
k =1A..JV
is the FT of v; (f), /*(©) is the FT of the short circuit current ik(t) at
port k, Ykj (co) is the mutual admittance between port k and j , and co is the angular
frequency. If Vj(co) = 1 for all co, then the short circuit current ik(t) in the time
domain will be the inverse FT of
Y kj
(co), that is,
(7.63)
where the subscript gkj denotes the Green’s function corresponding to the excitation
at port j and response at port k. If an arbitrary function Vj (co) is excited at port j ,
the response at port k will be
i k { t ) = F - l { Y kj { a y V j ( ( s » }
t
=
ig k j ( t )
*
Vj(t)
.
(7.64)
o
w h ere * d en o tes co n v o lu tio n . W hen all ports are ex cited , the cu rre n t at p o rt
w ritten as
k
can be
(7.65)
j =l o
If th e lo ad at p o rt
j
is lin ear, then there is a sim p le relatio n betw een Vy(co) and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 206Ij (co), and the current at each port can be obtained from the frequency domain by
solving the simultaneous equations. However, if the load is nonlinear, the relation
between the load current and the load voltage can no longer be expressed in fre­
quency domain, but has to be treated in time domain.
Let the v - i relation of the nonlinear load at port j be defined by
ij(t) = Fj(Vj(t))
(7.66)
where Fj is a known function. By substituting Eq. (7.66) into Eq. (7.65), one has
N
\
^ jt(v jk (0 )= E jigkj(t^t)Vj(x)dx
j=i o
(7.67)
The above equation can be expressed in discrete form, that is, the integration can be
replaced by the summation
Fk (vk (<?Af)) = £ £ igkj [(<7-p )&t]Vj (p At )At
;=1 p=0
*=1,..JV
(7.68)
where the subscript q denotes the q th instant time step. The sum in the above equa­
tion can be separated into two parts
Fk (v* (q At)) = '£ igkj(0 )Vj(qAt)&
j =i
+ £ qZ ig kj[(q-p)to]vj (pAt)At
j =1 p=o
k=\,...,N
(7.69)
The first term in the right part of the above equation contains the load voltage at
t=qAt, while the second term contains only the the load voltage at the previous
times.
The terms igkj (0) and igkj ((q -p )Ar)for agivenelectromagnetic
becalculated in advance. If the load voltageof the previous
system can
values are known, the
only unknown parameter in Eq. (7.69) is the instant ioad voltage Vj (q A t) at t=qAt.
This unknown can then be determined by solving the simultaneous nonlinear equa­
tions of Eq.(7.69) and the load currents can then be solved with Eq. (7.66).
Consider a scatterer with an n-port nonlinear load being illuminated by a plane
wave, its equivalent network can be expressed as shown in Fig. 7.17. Computation of
the radiated field requires an additional port to include the incident field as an
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
- 2 07-
r
N-port linear
network system
Yij{ G>)
»
1
8
i►
C*1
1
Fig.7.17 Equivalent network of an N-port nonlinear loaded scatterer.
excitation. The load current and the radiated fields can be calculated from the following equations
N q
N X
Fk (v* (qAt)) = £ £ igkj [(q - p )At ]vj (p At )At
i-\p = o
(p At )At
(7.70)
p=0
and
N
q
Erad (QAt) = X S
y=ip=o
TP
]vj(P & )Af
+ i ^ £ [(<7-p)Ar]EI„c(pA/)Ar
p=0
(7.71)
where igkE represents the short circuit current at port k due to an impulse incident
field excitation; EgJ and EgE represent the Green’s functions for the scattered field
when there is an impulse excitation applied at port j and an impulse incident electric
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 208 -
field respectively.
To reduce the required time duration of the impulse response at each port,
Djordjevic and Sarkar [99] proposed a method which involves inserting a suitable
resistance terminated at each port so that a shorter duration of Green’s function can
be used to carry out the calculation. To deduce the original characteristic, a negative
resistance of the same value is then connected in series with the inserted resistance.
7.2.2 Scattering Properties of a Nonlinearly Loaded Thin Rod Scatterer
Intuitively, it is expected that nonlinear loading will generate harmonics of the
incident wave, so that the energy of the fundamental mode of the back-scattered
wave will be reduced. In this subsection we will examine the above statement and
use a thin rod as a test object.
For a straight wire scatterer with given length to radius ratio (I/a), the time
functions ig^ (r), igkE(t), Egj(t), and EgE(t) of the previous subsection can be found
from the inverse FT of the frequency domain data which can be obtained by the
moment method as described in the previous section. These functions are independent
of the loaded impedance. The characteristics of the nonlinear loads are specified by
the v - 1 curve of the load. The frequently used nonlinear loads are the piecewise
linear resistors, or the loads with their v - i relation which can be expressed as a
summation of a power series.
T h e n o n lin e a r lo ad to be c o n sid ered is p iecew isely lin ear an d is c h ara cteriz ed by
(7.72)
w h ere v c u t is th e c u to ff v o ltag e o f the n o n lin ear load (fo r e x am p le, a diode),
y i a n d y 2 are th e fo rw ard an d rev ersed c o n d u ctan ce o f the load resp ectiv ely , and
v ( r ) an d t ( r ) are th e v o ltag e acro ss the load and th e cu rre n t flo w in g th ro u g h the load
re sp ec tiv e ly . In th e fo llo w in g calcu latio n s, w e w ill assu m e th at v c u t c a n alw ay s be
c o m p e n sa te d by a b ias v o ltag e so that the effectiv e v c u t is eq u al to zero. T h ere is an
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 209-
abrupt change in the slops when the load voltage passes through the zero value. This
abrupt change might generate a dc term and higher harmonics.
As seen from Eq. (7.71) the scattered field can be separated into two com­
ponents: a load-dependent part (t!<s first term) and a load-independent part (the
second term). If the illuminated wave is a monochromatic plane wave, the loadindependent component is the field scattered from the load-free scatterer and is
sinusoidally varied in the steady state. The nonlinear load only affects the first term
of Eq. (7.71). In the following examples we will discuss these two components indi­
vidually.
Consider a thin rod, with length equal to three wavelengths and a length to
radius ratio equal to 100, that is illuminated by a sinusoidal plane wave from broad­
side. A lead is inserted at the center of the rod. If the characteristic of the load is
piecewisely linear with y i = 1 and y 2 - 0.01, the time domain back-scattered field of
the load-dependent part is shown in Fig. 7.18(a). Its spectrum is shown in Fig.
7.18(b), which is obtained by a FT of the time domain field. It is seen that addi­
tional harmonics appear in the spectral plots. If the load is linear with y i = y 2 = 1,
its corresponding time-domain field and spectra are shown in Figs. 7.18(c) and
7.18(d) respectively. While shown in Figs. 7.18(e) and 7.18(f) are the counterparts
when the load is linear with y j = y 2 = 0.01. Note that the scales in the above figures
are not all the same. In reference to the above spectral plots, one can see that the
magnitude of the fundamental mode with nonlinear loading is greater than that with
_yj = jy2 = 1 but is smaller than that with yj = y 2 = 0.01. From the viewpoint of
RCS reduction, fixed nonlinear loading is ineffective at reducing the RCS at a certain
frequency because one usually can have a smaller RCS by suitably choosing a linear
load with conductance equal to one of the two slopes of the v - i curve. The loadindependent component of the time domain field is shown in Fig. 7.18(g). It is seen
that the strength of the load-independent part is much stronger than that of the loaddependent part. The spectra of the back-scattered field before and after nonlinear
loading are shown in Figs. 7.18(h) and (i). The above figures tell us that a single
nonlinear load has very little effect on the RCS and spectra when the wave is
illuminated from the broadside direction.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-210-.
l.MOOs
•
— ' c a n p la p o i n t
fig .7 .1 8 (a)
•-06
f i g . 7 . 16(b )
V
>
M
1
I
t
ud
a
(
er
•
I
\
t
1.3225a*02
aampla p o i n t
fig.?.l8<«?
V
a
"
f
1.8088
«-»7
1
t
)) 1.2m
;
.-97
’\f*jwywv\<
at 7.18(d)
7 jg Option see p3ge 212)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-211 -
9
m
T
l
t
u
d
9
t
r
•
a
1
t.
lanpl*
♦'tg7.ia(9>
—-
po in t
a
n
?
i
t
u
d
•
Fig. 7.18 (caption see next page)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-212-
1.49994*99
•
—
l a x s l a p o in t
flj.7 .t9 !3 )
2.9449
4-94
1.3999
-
■
4 -9 *
1.9494
4-94
9 . 9999
« -« 9
k.
'
-7.34944*46
fij.
-3 .7 S 4 « 4 * 9 9
-2 .92974-92
3.79914*99
(h)
7 .1 9 (h )
2 .9 9 9 9
4-64
" 1.3969 -■
f
4-94
1
,
t
i
I
1 .9 9 9 9
4 -9 4
3 .9 9 9 9
4-99
s-7.59944*49
♦'i j. 7 , t 0 ( i )
1
Jv-
- 2 .9 2 9 7 « - * 2
3 .7 9 9 1 4 * 9 9
(i)
F ig .7 .1 8 T im e d om ain field s and their spectra o f a n on lin early and lin early loaded
w ire scatterer. (a) T h e load -d ep en d en t p a n o f the back -scattered field o f a non­
lin early lo a d ed w ire w ith y j = 1 a n d y 2 = 0 .0 1 . (b) Sp ectral p lot o f (a), (c)
T h e lo a d -d ep en d en t p a n o f the back -scattered field o f a n on lin early lo a d ed w ire
w ith y j = y 2 = 1* (d ) Spectral p lo t o f (c). (e ) T h e lo a d -d ep en d en t p a n o f the
b ack-scattered field o f a n onlinearly lo a d ed w ir e w ith y l = y 2 = 0 .0 1 . (f) S p ec ­
tral p lo t o f (e). (g ) T h e lo a d -in d ep en d en t p a n o f the back -scattered field o f a
thin w ire, (h ) Sp ectral p lot o f (g ). (i) T h e spectra o f the back -scattered field o f
a n o n lin early lo a d ed thin w ire w ith y i = I and y 2 = 0 .0 1 .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 213 -
i
1 . mi»*ii
—
:
V
a
"
If
i
t
(
s a x p l a p o in t
4.0000
e-0 4
3.0000
e-0 4
'i 2. m i
;
a-0 4
1.0000
a-0 4
-
•
-----------------ji -----------'
i
Fig.7.19 Time domain scattered fields and spectra of a nonlinearly loaded wire
scatterer with three loading points, (a) The load-dependent component of the
back-scattered fields, (b) The spectra of the total back-scattered field.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 214-
If we increase the number of point from 1 to three and locate them at
z =0, ±1 / 2 , the load-dependent component of the time domain field and the resultant
spectra of the total back-scattered field are shown in Figs. 7.19(a) and (b) respec­
tively. Compare Fig. 7.19(a) and Fig. 7.19(b), one can see that increasing the
number of loading point increases the contribution of the load-dependent component.
However, the effect is still very small with broadside illumination.
7.3 Time Varying Loading
The loads considered in the previous sections are fixed loads. In this section we
will consider another type of load, time varying load, whose impedance value is time
dependent Advances in technology have made it possible to control the load elec­
tronically. It has been pointed out that a time varying load might shift the apparent
frequency of the scattered field to provide a false doppler shift and to make the
receiver at the probing radar unable to phase lock to the frequency of the incident
wave [85]. An example of a half-wave dipole scatterer with time varying resistive
loading was given in [85]. In this section we will study some other interesting
phenomena caused by a time varying load and will numerically examine how
effective the time varying load may be in producing the expected effects. The object
to be tested is a thin rod and the factors to be considered are the type of loads, length
of rod, and direction of incident waves.
An incident scalar wave with angular frequency co0 and magnitude £,° can be
expressed as
£ ,(r ) = R e t E . V 0* }
(7.73)
where Re{*} denotes the real part of *. The doppler frequency corf of an object
moving with velocity u toward the radar is then
u
v>d = “ o—
c
(7.74)
where c is the light velocity. Assume that the load value is varied with time and is
denoted by ZL(t), the time domain back-scattered field is then
Es (0 = Re { E's (t)e )(o)o+(OdX
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(7.75)
- 215 -
where
E's (t) = Es [ZL (t)]
(7.76)
If ZL (t) is a constant, the doppler frequency which appears in radar is (ad . If ZL(t) is
sinusoidally varied with frequency cot, the spectra of the scattered field will have the
following components: co0 + 0)rf ± «<»!, where n is an integer. However, the magni­
tude of each spectral component depends on how E's [ZL(t)\ changes with respect to
ZL(t). If the variation in the magnitude and phase of E's [ZL (t)\ is great during
excursion of one cycle, the portions of the spectral components other than ©0+0)d
will be more pronounced, which may provide the radar with a false doppler fre­
quency. On the other hand, if the variation is very small, the other spectral com­
ponents will be much smaller than the dominant component (Oq+co^. In that case,
time varying loading is ineffective at providing false doppler frequencies.
The effectiveness of a time varying load can be measured by a modulation
index defined as
_ \E's {ZL(t))\ max
11 “ IE's (ZL (t)) Imax
IE's (ZL(t))\
min
(7.77)
\E’s (ZL(t))\ min
The value of T| is between 0 and 1. If the modulation is close to 1, the scheme will
be more successful in providing a false doppler shift. If T| is very small, the scheme
cannot provide the desired effect.
In Fig. 7.5, we have plotted the possible maximum and minimum values of the
back-scattered field of a central passively loaded straight wire versus frequency at
certain aspect angles. From those plots, we can predict how effective the time vary­
ing loading is.
In the following we will examine the spectra of Es (t) of a sinusoidally varied
load, either varied resistively or reactively. Assume the resistively varied load is
given by
ZL(t) = RL{t) = 250(1-coscojt) Q
while the reactive loading is given by
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(7.78)
- 216 -
ZL (t)= j XL (t) = j 250 sinov Q
(7.79)
the loading point is at the center of the load. Shown in Fig. 7.20 are the spectra of
E's(t) of a straight wire operated at several wavelengths and incident angles. The
modulating index of each case is also shown in the accompanying figures. From Fig.
7.5 and Fig. 20, we make the following conclusions:
1.
There are more sidebands in the spectral plots when the modulation index is
greater.
2.
Reactive loading usually has a greater modulation index than resistive loading
does, which implies that reactively time varying load are more effective at pro­
ducing a false doppler shift.
3.
The modulation index is small in the broadside case when the length of the wire
is greater than 1.4 wavelengths.
4.
In the side-lobe regions, the modulation index can be close to 1 for many fre­
quencies but not for every frequency.
The above analysis and conclusions are examined from the consideration of pro­
viding a "false" Doppler frequency. However, the same conclusions can also be
applied in the RCS consideration. A radar receiver usually has a bandwidth limita­
tion. If the load varying frequency is greater than the bandwidth of the receiver, the
sideband energy will be rejected by the receiver. In this sense we can say that time
varying loading does spread the spectrum of the back-scattered field. It is noted that
the philosophy of diaphanization by nonlinear loading is also based on the spectrum
spreading effect. Nevertheless, time varying loading differs from nonlinear loading
in the following:
a.
The separation between the successive harmonics in time varying loading is the
load varying frequency, while the separation between the successive harmonics
in nonlinear loading is the frequency of the incident wave.
b.
When the modulating frequency of the time-varying load is much smaller than
the carrier frequency of the incident wave, time-varying loading can be viewed
as a form of linear loading. In that case time-varying loading can be analyzed in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- LIZ-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
F ig .7 .2 0 Spectra o f a tim e vary in g load ed scatterer w ith param eters
(a) /=1.2X, 0 = 90°, T] = 0.21, nL =
(b) / = \.2X, 0 = 90°, n = 0.82, nL =
(c) / = 3 . OX, 0 = 90°, T| = 0.24, nL (d ) / =3.0X., 0 = 4 5 ° , t) = 1.0, nL = 1,
1, z= 0, resistive loading.
1, z= 0, reactive loading.
1, z = 0, reactive loading.
z = 0 , rea ctiv e load in g.
- 218 -
the frequency domain. While nonlinear loading can only be analyzed in the
time domain.
Practically, it is much easier to produce a time varying loading effect than to
produce a nonlinear loading effect in the application of radar scattering. An input
bias energy (cutoff energy) is usually required to make a practical diode ( for exam­
ple, a step recovery diode) producing the desired nonlinear property. However, the
incident wave energy is usually not great enough to satisfy this requirement. On the
contrary, the input impedance at a loading port can be easily varied. For example,
the input impedance at the loading port can be electronically controlled by inserting a
digital phase shifter or a tuning stub between the loading port and the terminating
loaa.
We will now examine the effect of time varying loading on the range profiles
and the reconstructed images. The loaded values of the examples shown in Sec.
7.1.3 are fixed for each frequency and each aspect angle. Examining the range
profiles of those examples and the range profiles shown in chap.6, one can find that
there are more peaks in the range profiles due to the created discontinuities. How­
ever, those peaks which are a result of reflection from the end points of the object
can still be easily observed in the range profile even though some extra peaks have
been created between them. The range distances at the right most and the left most
peaks in the range profile usually provide information about the object’s dimension.
One may use an active broad-band slave jammer to obscure or to distort the range
information, but this is not what we wish to discuss. We try to use a passive
impedance load to. achieve this goal. Impedance loading can change the magnitude
and phase of the scattered fields. If the loaded values are randomly varied for each
loading point, each frequency, and each aspect angle, that is. if they are randomly
varied for each time instant, this randomness might cause random peaks in the range
profile. However, if the number of loading points is too small, or if the range for
which the impedance value is varied is not suitably chosen, the random loading may
not give the desired effects.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 219 -
We have concluded that reactive loading can make a more drastic change in the
scattered fields (either in phase or amplitude) than resistively loading can, and
increasing the number of loading points can produce greater field variation. It is also
known that the reflection coefficient at a given point is a function of the input
impedance and the loading impedance at that point. If the loaded impedance at a
point is randomly switched between capacitive loading and inductive loading, the
phase of the reflection coefficient at that point will be changed at each time instant.
Consequently, the range profile may produce more random peaks and the magnitude
of the peaks corresponding to the loading point may be reduced.
In the following we will compare the effect of a fixed loading, a randomly resis­
tive loading, and a randomly reactive loading on the back-scattered field and the
range profiles of a loaded straight wire. The parameters used are nL=5, and 6=45°.
Each loading impedance is either fixed to SO ohms, or randomly resistively varied
from 0 to 100 ohms, or randomly reactively from -j50 ohms to +j50 ohms. The
magnitude and phase of the back-scattered field and the range profiles of the above
three loading cases are shown in Figs. 7.21, 7.22, and 7.23 respectively. From the
above figures one can find that the difference in the RCS with fixed resistive loading
versus that with randomly resistive loading is small and randomly resistive loading
does not create random peaks in the range profile. If the number of loading point is
changed to 3 and 7 and each load is randomly varied between (-y'50, y'50)Q, the
corresponding range profiles are shown in Figs. 7.24(a) and (b) respectively. It is
seen that a small number of loading point is inadequate for generating random peaks
and additional loading points obscure the range profile more effectively.
The real part and imaginary part of the back-scattered fields and the recon­
structed images of the randomly reactively loaded wire with nL=7 are shown in Fig.
7.25. Comparing Fig. 7.25 with Fig. 7.17, one sees more noise in the image of the
ranH nm lv rpnpHvpIv InaHpH sraftprpr
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 220
-
.75
.50
o.
rm
3
.25*
1.1
4.1
10.0
7.0
length in terms of wavelength
(a)
13.0
180
-180 QL
1.1
u
4.1
7.0
-37
0
distance in cm
(c)
13.0
10.0
length in terms of wavelength
(b)
.075
-3
3
a
.050
.025 -
-74
74
37
F ig -7 21 T h e (a) m a g n itu d e and (b ) p h a se o f the b a c k -sca ttered field , a n d (c ) the
ra n g e
nL= 5, z
p r o file
of
the
lo a d ed
straigh t
w ire
[ = 0 .6 7 / = - z 5 z 2= 0 .3 3 b - z 4, z 3= 0 , an d 0 = 4 5 °
w ith
zL =50Q,
w h ere
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 221
-
a -75
-3
! , .5 0
1.1
4.1
10.0
13.0
length in terms of wavelength
7.0
(a)
180
00
a
"O
c
-180
1.1
4.1
10.0
7.0
13.0
length in terms of wavelength
(b)
a
-3
3
Q.
s3
.075 -
.0 5 0
.025
-74
-37
0
distance in cm
37
74
(c)
F ig .7 .2 2 T h e (a ) m a g n itu d e and (b ) p h a se o f the b a ck -sc a tter e d fie ld , an d (c ) the
ran ge p ro file o f th e lo a d ed straigh t w ire, e a c h im p e d a n c e is ran d om ly r e sistiv e ly
v a ried
b e tw e e n
(0, 100)Q, w h ere nL= 5, z l=Q.67l=-2S z2=0.33/»-Z4,
z3=0 , and 0=45°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 222
«
*3
3
-
.75
9* .50
.25
l.i
4.1
7.0
10.0
length in terms of wavelength
(a)
13.0
180
ao
u
•a
e
-180
1.1
4.1
7.0
10.0
length in terms of wavelength
13.0
(b)
.075
i . .050
.025
-74
-37
0
distance in cm
37
74
(c)
Fig.7.23 T he (a) m agnitude and (b) phase o f the back-scattered field, and (c) the
range profile o f the loaded straight wire, each im pedance is reactively random ly
varied betw een (-y 50, y 50)Q , where r t L - 5 , z l= 0 .6 7 /= -?5 i z 2=0.33/>Z4,
z 3=0, and 0=45°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 223 -
v
a
n
P
1
7.5 000
e-0 2
i
t
u
a
e
— > d i s t a n c e i n cm
fig .
7 . 24 (a)
(a)
1.0000
e-01
7 .5 0 9 9
e-0 2
5 .0 0 9 0
a-0 2
2 .5 0 0 9
a-0 2
-7.5 0 0 9 a* 0 1
- 3 . 7646e+01
—
d is t a n c e in cm
-2 .3 2 9 7 e -0 1
...
fig.
'
7 . 2 4( b )
3 . 706ie*01
^
Fig.7.24 The range profiles of a randomly loaded straight wire with
(a.) nL = 3, z x = 0.5/ = - z 3, and z2=0.
(b) nL = 7 , z { = 0.951 = - z 7, z2 = 0.67/ = - z 6, z3 = 0.33/ = - z 5, and z4 = 0.
Each impedance is randomly varied between (—y 50, y‘50)Q, where 0 = 45°.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 224 -
d 1p r n n .7 r ; NL=7
(.96,-1.75);(.5,-.05)
d l p r a n . 7 1 ; N_=7
(4.81,-3.32);(.5,-05)
I
(a)
(b)
iipoinfLB^jjfTseasp) i (ge.oo
d y n a (lS 7 .0 )i(1 5 0 ,0 )'
(C)
'
(d)
Fig.7.25 (a) Real part and (b) imaginary part of the range-corrected scattered field
of the randomly reactively loaded straight wire with nL = 7, z. - 0=95/ = - z 7.
z2 = 0.67/ = -Zf,. z3 = 0.33/ = - z 5, and z4 = 0.
The reconstructed image using angular window over (c) (20°, 90°) and (d) (20°,
84°).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 225 -
7.4 Discussion
A thin rod has been used as a test object throughout this chapter to numerically
examine how linear loading, nonlinear loading, and time varying loading affect the
scattering properties and images of an object A scale model Boeing 747 aircraft has
been approximated by a pipe model and a thin wire model and their transient
responses have been measured and compared [38]. In this section we will first
present microwave images of a wire model airplane and then discuss the acceptability
of a thin wire in the approximation of a finite edge.
An optical image and a microwave image of a wire model airplane are shown in
Fig. 7.26(a). The object was mounted on a rotating pedestal with inclination angle
equal to zero degrees. The angular window used is from broadside to nose-on and
the frequency coverage is from 6 GHz to 16.5 GHz. Examining the reconstructed
images one can find that the image intensity of the front "edge" and rear "edge" of
the right "wing" is almost of equal brightness. The above phenomena are the major
differences from those of a scale model as shown in the previous chapters. The rea­
son for this difference may be explained as follows: most parts of this wire model
object can be illuminated by the incident wave. In other words, no shadow region
exists in this object.
Next we will discuss the acceptability of a thin wire in the approximation of a
finite edge. Consider a z directed wedge with finite length L as shown in Fig. 2.8. It
has been pointed cut that the back-scattered field of an edge is a function of the
azimuth angle and the elevation angle. The back-scattered field pattern is a sine
function along the 0 direction and is a function of <|> in the azimuthal direction. In
contrast, the back-scattered field pattern of a long wire is close to a sine function
(except near the end-on direction) along the 0 direction and is isotropic in the <)>
direction. If we divide the wedge into two regions by introducing an intervening
insulator as shown in Fig. 6.27, the back-scattered field can be expressed as the
superposition of the fields scattered from the narrow strip and the remaining plates.
The above statements assumes that the narrow strip and the remaining plates are iso­
lated scatterers (i.e., no mutual coupling existing between them) and the scattering
from the insulator is much smaller than that from the conducting parts because of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(a)
(b)
Fig. 7.26 Optical image and microwave image of a wire model airplane
(a) optical image
(b) Co-polarized image reconstructed over angular window (<J> = 0° to <j) = 90°)
perfect conductor
insulator
Fig. 7.27 A loaded conducting wedge.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 227 -
low reflectivity of the insulator. The narrow strip might then be approximated by a
thin wire and the properties of a loaded thin wire examined in the previous sections
may be applied to the narrow strip case.
As indicated in Sec. 6.4, the major contributors to the scattered field of a plate
are their tips when the incident wave is not normal to the edges and the equivalent
scattering strength from a tip at a given aspect is not a function of the length of the
edges. Therefore, the scattered field from the narrow strip is comparable to that from
the remaining plates when the aspect is in the side-lobe region. For the above reason
it is expected that the effect of introducing impedance loading to the narrow strip
may not be overridden by the presence of the remaining plates.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 228 -
CHAPTER 8
CONCLUSION
In this dissertation we briefly review the basic scattering properties of a perfectly
conducting object. In the high frequency region, the scattered field of a complex
shaped object can be attributed to a combination of several mechanisms. The advent
of high resolution radar enables us to consider each scattering mechanism separately.
We have interpreted the microwave image of a conducting object from a new view
point, based on the understanding of the scattering mechanism and the image recon­
struction algorithm. The connection between various scattering mechanisms and their
reconstructed images is then established, from which we can interpret what the image
represents and predict what the image will look like over given spectral and angular
windows. Several numerical and experimental examples have been included to sup­
port this new approach to image interpretation.
Based on the understanding of the scattering mechanism and the knowledge of
modem spectral estimation techniques, we devise a new algorithm to extrapolate the
available data into the missing band and apply this algorithm to radar imaging.
Both simulation and experimental results have shown the effectiveness of this method
in microwave diversity radar imaging. A significant application of this new algo­
rithm is in identifying the hot spots of a target when the bandwidth of the imaging
system is too narrow to give acceptable resolution.
The satisfactory interpretation of microwave images and the effectiveness of the
extrapolation algorithm devised are fundamental to the study of RCS management
and target recognition. We introduce a new term, "diaphanization", defined as the
techniques of reducing RCS and the techniques of obscuring an image. We consider
the RCS management not only from a detection perspective, the reduction of the
target’s RCS to elude radar detection, but also from the image point of view, the dis­
guising of a target’s appearance to impede radar recognition. This consideration
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 229-
opens a new direction for future research in RCS management studies.
We give the procedure which employs microwave diversity imaging to diaphanize a target over prescribed spectral and angular windows and demonstrate the robust­
ness of the diversity imaging system to the Gaussian noise. We also propose some
rules for distorting an image. These rules are: create artificial discontinuities, create
multiple reflections and make the reflectivity at a given point a function of time.
Traditional techniques for reducing the RCS are applied to RCS management
studies. These techniques are absorber covering, target shaping, and impedance load­
ing. We review the theory pertaining to absorber-covered bodies and establish the
theoretical background for obscuring an image by absorber covering based on the
physical optics approximation. Certain types of absorbers are used to cover metallic
objects. Their influence on RCS reduction, range profile modification and image dis­
tortion are experimentally studied and discussed. We propose a concept concerning
the absorber covering patterns where we also consider the reflections from the boun­
daries between the covering absorbers and conductors. By using suitable covering
patterns the RCS can be reduced more effectively.
Finally we examine some possible effects of loading an object with lumped
impedances. A thin straight wire is used as a test object. Plots of range profiles
show that the surface traveling wave is an important scattering mechanism of a
straight wire. Microwave images of a loaded wire are both numerically and experi­
mentally obtained. The specific scattering mechanism of the thin wire makes its
image unique in appearance. We also demonstrate that nonlinear loading and time
varying loading can cause the spectra of the back-scattered field to spread from the
incident wave, which may produce several interesting phenomena, such as providing
a false doppler frequency, making the receiver unable to phase lock to the frequency
of the incident wave, generating unexpected peaks in the range profiles, and distort­
ing the reconstructed image.
The main imaging geometry of concern in this dissertation involves rotation of
objects. However, the approach to image interpretation described in Chap.3 can also
be applied to bistatic angular diversity imaging, which is to synthesize the angular
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 230-
aperture by varying the bistatic angle between the transmitter and receiver. One may
first study the bistatic scattering properties and range profiles for various scattering
mechanisms and then back-project the range profiles according to the relative motion
between the transmitter and receiver. Because the information collected by bistatic
diversity imaging is different from that by rotating an object, the reconstructed
images are expected to be different. However, it is possible to predict the image
from the sinogram of the bistatic image.
8.1 Future Research
Several areas which may be further investigated are suggested below:
A. Analysis of microwave diversity imaging constructed in this dissertation
assumes that the observation distance is in the far field region. The scale model
airplane used is 1:100. Assume the frequency operated in the real system is 2
GHz, then the frequency in the scale system should be 200 GHz. Therefore,
millimeter wave imaging is more desirable in order to simulate a real system.
However, the observation distance conducted in the anechoic chamber will be in
the near field region for such high frequencies. A thorough analysis of near
field diversity imaging for a conducting object has not been reported and should
prove to be a challenging research topic.
B.
Interpretation of the microwave images conducted in this dissertation also
assumes that the transmitter and receiver both lie in the azimuthal plane perpen­
dicular to the rotational axis. The image reconstructed is a projective image. If
the line of sight is not perpendicular to the rotational axis, the image will be
unfocused. Seeking an efficient algorithm to correct the unfocused image would
be an interesting task.
C.
The range resolution of microwave diversity imaging depends on the bandwidth
available while the cross-range resolution depends on the angular aperture in
terms of wavelength. The extrapolation algorithm developed in Chap.4 is to
extrapolate the scattered fields into the exterior band for each fixed aspect. The
applicability of this algorithm is based on the fact that the scattering properties
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 231 -
of each scattering mechanism are independent or only slightly varied with
respect to frequency. In practice the angular window available may be too nar­
row to give acceptable cross-range resolution. Methods of increasing the crossrange resolution are worth investigating.
D.
The images obtained are projective images. The locations of the hot spots in
the direction of the rotational axis are still ambiguous. A 3-d tomographic
image can pinpoint the hot spots more accurately. Analysis of 3-d tomographic
images for conducting objects was based on the physical optics approximation
[9]. However, that interpretation encounters the same difficulties as mentioned
in Chap. 3. Therefore, a more convincing interpretation for 3-d imaging of a
complex shaped metallic object needs to be developed.
E.
The comparison of the images before and after diaphanization are made by the
human brain. That judgement is subjective. How to combine machine vision
with knowledge of scattering and image formation should prove to be very valu­
able future research work. Automatic machine recognition and evaluation of
microwave diversity data based on models of neural networks is an active
research area at the Electro-Optics and Microwave Optics Laboratory.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 232-
APPENDIX
Consider two identical parallel cylinders with radius a, separated with a distance
2r 0 seated on a rotating pedestal and illuminated by a plane wave as shown in Fig.
5(a). The cross section of that arrangement is shown in Fig. Al. The line connecting
the centers of the cylinders passes through the rotating center and makes an angle 6
with the respect to the rotation axis.
The polar coordinatesandrectangular coordi­
nates of the points on the surfaces of these two cylinders with respect to
cylinder center are (a ,
each
y ^ , and (a, <j>2'), (x2f, y { ) respectively. Let the bis­
tatic angle between the transmitter and receiver be a, the unit vectors in the direction
of transmitter and receiver be I, and lr respectively. To find the stationary points of
the multiple reflection, geometrical optics will be applied. The reflection law requires
that the angle between the incident ray and normal fine must be equal to the angle
between the reflected ray and the normal line.
The stationary points of multiple reflection on the respective cylindrical surfaces
for a given set of {r0, a , 6, a} are to be determined. Assume the incident ray first
hit cylinder 2, reflect to cylinder 1, and then bounce back to the receiver as shown in
Fig. Al(a). The notation of the angles I;,
i|f2, a, 0, 0i',
§2
*s defined in Fig.
A 1(a). It can be shown the following relationship must be satisfied in order to
satisfy the reflection law.
2 rosin0 - a [costo?' - aJ2 ) + sin(W]
— H-----------.— ILf----------------- — = -cot(20, - a/2 )
2 rocos0 + a [sin(<p2' - a/2) -cos<)>2 ]
(A.l)
<t)j' = q>2' - 90° - a/2
(A.2)
where <|>2' is restricted to be
270° < 02 ^ 360° when
-90° < 0 < 90°
180° < 02' < 270° when
90° < 6 < 270°
For a given set of parameters {r0, a, 0, a }, the azimuth angle <t>2' and therefore 0i'
can be determined by solving Eq.(A.l) and (A.2).
The other case is that the incident ray first hits cylinder 1, reflects to cylinder 2,
and then bounces back to the receiver. The geometry of this situation is shown in
Fig. A 1(b). The stationary points must satisfy the following relationship
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-2 3 3
y
012
a ll
y
y
a ll
Fig.Al Multiple reflections of a two-cylinder object, (a). The incident wave reaches
the left cylinder first, and is then reflected back by the right cylinder, (b). The
incident wave reaches the right cylinder first and is then reflected back by the left
cylinder.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 234 -
<J>i" = <>2" - 90° + a/2
2 rosin0 - a [cos( <{>2" + a/2) + sin<|>2"]
2rocos0 + a [sin( <|>2" + a/2 ) - cos<t>2"]
(A.3)
= -cot(2<|>2" + a/2)
(A.4)
From the above derivation, it can be seen that there are two pairs of multiple
reflection points for a bistatic system and only one multiple reflection point for the
monostatic case.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 235 -
BIBLIOGRAPHY
[1]
H. P. Baltes, ed., Inverse Scattering Problems in Optics, Springer-Verlag, New
York, 1980.
[2]
W-M. Boemer, "Polarization Utilization in Electromagnetic Inverse Scattering,"
in Chapter 7 of Inverse Scattering Problems in Optics, H. P. Baltes ed.,
Spring-Verlag, New York, 1980.
[3]
L. J. Cutrona, E. N. Leith, L. J. Porcello, and W. E. Vivian, "On the Applica­
tion of Coherent Optical Processing Techniques to Synthetic Aperture Radar,"
Proc. IEEE, Vol. 54, No. 8, pp. 1026-1033, 1966.
[4]
Kiyo Tomiyasu, "Tutorial Review of Synthetic Aperture Radar (SAR) with
Applications to Imaging of Ocean Surface," Proc. IEEE, Vol. 63, No. 5,
pp.563-583, 1975.
[5]
N. H. Farhat and C. K. Chan, "Three Dimensional Imaging by Wave-Vector
Diversity," in Acoustical Imaging, Vol. 8, Edited by A. F. Metherell, Plenum
Press, New York, pp. 499-516,1980.
[6]
J. L. Walker, "Range-Doppler Imaging of Rotating Objects," IEEE Trans.
Aerospace and Electronic Systems Vol. AES-16, Vol. 1, pp. 23-52, 1980.
[7]
D. C. Munson Jr., J. D. O’brien, and W. K. Jenkins, "A Tormographic Formu­
lation of Spotlight-mode Synthetic Aperture Radar," Proc. IEEE, Vol. 71, No.
8, pp.917-925, 1983.
[8]
N. H. Farhat, "Principles of Broad-Band Coherent Imaging," J. Opt. Soc. Am.,
Vol. 67, No.8, pp. 1015-1021, 1977.
[9]
C. K. Chan, and N. H. Farhat, "Frequency Swept Tomographic Imaging of
Three-Dimensional Perfectly Conducting Object," IEEE Trans. Antenna and
Propaga., Vol. AP-29, pp. 312-319, 1981.
[10] N. H. Farhat, C. L. Wemer and T. H. Chu, "Prospect for Three-Dimensional
Projective and Tomographic Imaging Radar Networks," Radio Science, Vol. 19,
No. 5, pp. 1347-1355, 1984.
[11] T. H. Chu, "Optimal Methologies in Inverse Scattering Utilizing Wavelength,
Polarization and Angular Diversity," Ph.D. Dissertation, University of Pennsyl­
vania, Dec. 1983.
[12] Yuhsyen Shen, "Dynamic Microwave Imaging," Ph.D. dissertation, University
of Pennsylvania, 1987.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 236-
[13] G. Tricoles and N. H. Farhat, "Microwave Holography: Applications and Tech­
niques," Proc. o f the IEEE, Vol. 65, No. 1, pp. 108-121, 1977.
[14] K. Iizuka and L. G. Gregoris, "Application of Microwave Holography in the
Study of the Field from a Radiating Source," Applied Physics Letters, vol. 17,
No. 112, 15 Dec. 1970.
[15] T. Sato, O. Ikeda, and K. Endo, "Combined Spectral and Aperture Synthetic
Ultrasonic Imaging System," IEEE. Trans. Sonics and Ultrasonics, Vol. SU28, pp. 64-69, 1981.
[16] H. Emert, and R. Karg, "Multifrequency Acoustical Holography," IEEE.
Trrans. Sonics and Ultrasonics, Vol. SU-26, pp. 279-286, 1979.
[17] H. J. Li, H. A. Yeuh, and G. Wade, "Using Multiple Frequency Method for
Artifact Reduction," Proc. o f the National Science Council, Taipei, Republic of
China, Part A: Applied Science, Vol. 7, No. 4., pp.301-312, O ct, 1983.
[18] H. J. Li, "Image Reconstruction of Microwave Tomogram," Journal of the
Chinese Institute o f Engineering, Taipei, Republic of China, Vol. 7, No. 4,
pp.255-264, 1984.
[19] R. G. Reeves, e t al., ed., Manual o f Remote Sensing, American Society of
Photogrammetry, Falls Church, Virginia, 1975.
[20] R. G. Reeves, et. al., ed., Manual o f Remote Sensing, Chap. 14, "Fundamentals
of Image Interpretations," American Society of Photogrammetry, Falls Church,
Virginia, 1975.
[21] G. T. Ruck, D. E. Barrick, W. D. Stauart, and C. K. Krichbaum, Radar Cross
Section Handbook, Plenum Press, New York, 1970.
[22] N. Boijarski, "Inverse Scattering," Final Rep. N000-19-73-C-0312F, Nov. Air
Syst. Command, Waiminister, Pa., Feb. 1974.
[23] R. M. Lewis, "Physical Optics Inverse Diffraction," IEEE Trans, on Antenna
and Propag., Vol. AP-17, pp. 308-314, 1969.
[24] M. R. Wohlers, S. Hsiao, J. Mendelsohn, and G. Gardner, "Computer Simula­
tion of Synthetic Aperture Radar Images of Three-Dimensional Objects," IEEE
Trans, on Aerospace and Electronic Systems, Vol. AES-16, pp. 258-271, 1980.
[25] G. T. Herman, ed., Image Reconstruction from Projections, New York:
Springer-Verlag, 1979.
[26] G. T. Herman, Image Reconstruction from Projections, New York, Academic
Press, 1980.
[27] R. M. Mersereau and A. V. Oppenheim, "Digital Reconstruction of Multi­
dimensional Signals from Their Projections," Proc. IEEE., Vol. 62, pp. 1319-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 237 1338, 1974.
[28] H. H. Stark, J. W. Woods, I. Paul, and R. Hingorani, "Direct Fourier Recon­
struction in Computer Tomography," IEEE Trans. Acoust., Speech, Signal Pro­
cess., Vol. ASSP-29, pp.237-245, 1981.
[29] N. H. Farhat and T. H. Chu, "Tomography and Inverse Scattering," Interna­
tional Commision on Optics, ICO-13, C l, pp. 62-63, 1984.
[30] C. C. Chen and H. C. Andrews, "Multifrequency Imaging of Radar Turntable
Data," IEEE Trans, on Aerospace and Electronic Systems, Vol. AES-16,
pp. 15-22, 1980.
[31] D. Mensa, G. Heidbreder, and G. Wade, "Aperture Synthesis by Object Rota­
tion in Coherent Imaging," IEEE. Trans. Nucl. Sci., Vol. NS-27, pp.989-998,
1980.
[32] D. Mensa, G. Heidbreder, and G. Wade, "Coherent Doppler Tomography for
Microwave Imaging," Proc. IEEE, Vol. 71, No.2, pp.254-261, 1983.
[33] E. F. Knott, "Radar Cross Section Reduction", in Chap.7 of Radar Cross Sec­
tion, E. F. Knott, ed., Artech House, Dedham, MA. 1985.
[34] N. H. Farhat, class notes for the course of Microwave Imaging, Holography
and Synthetic Aperture Systems, University of Pennsylvania.
[35] R. F. Harrington, Field Computation by Moment Methods, The Macmillan
Company, New York, 1968.
[36] J. Moore and R. Pizer, ’Moment Methods in Electromagnetics’, Research Stu­
dies Press LTD. Letchworth, England, 1984
[37] E. K. Miller and A. J. Poggio, "Moment-Method Techniques in Electromagnet­
ics from an Application Viewpoint," in Electromagnetic Scattering, P. L. E.
Uslenghi ed., Academic Press, New York, 1978.
[38] E. K. Miller and J. A. Landt, "Direct Time-Domain Techniques for Transient
Radiation and Scattering from Wires," Proc. IEEE., Vol. 68, pp. 1396-1423,
1980.
[39] R. Mittra, "Integral Equation Methods for Transient Scattering," in Chap.2 of
Transient Electromagnetics, L. Felsen, ed., New York, Springer Verlag, 1976.
[40] E. K. Miller and F. J. Deadrick, "Some Computational Aspects of Thin-wire
Modeling " in Numerical and Asymptotic Techniques in Electromagnetics, R.
Mittra, ed., Springer-Verlag, New York, 1973.
[41] R. G. Kouyoumjian and P. H. Pathak, "A Uniform Geometrical Theory of
Diffraction for an Edge in a Perfectly Conducting Surface," Proc. IEEE., Vol.
62, pp. 1448-1461, 1974.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 238 -
[42] M. Bom and E. Wolf, Principle of Optics 6th ed., Fergamon, Oxford, 1980.
[43] R. S. Elliot, Antenna Theory and Design, Prentice-Hall, Inc., Englewood, New
Jersy, 1981.
[44] W. L. Stutzman and G.A. Thiele, ’Antenna Theory and Design’, Chap. 7, John
Wiley & Sons, Inc. 1981.
[45] J. B. Keller, "Geometrical Theory of Diffraction," J. Optical Society o f Ameri­
can, Vol.52, No.2, pp.l 16-130, 1962.
[46] R.. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New
York, 1961.
[47] E. F. Knott, "High Frequency RCS Prediction Techniques," in Chap. 5 of
Radar Cross Section, edited by E. F. Knot e t al, Artech House, Dedham, MA.
1985.
[48] A. Michaeli, "Equivalent Edge Current for Arbitrary Aspects of Observation,"
IEEE Trans, on Antenna and Propaga.d, vol AP-32, No. 3, pp. 252-258, 1984.
[49] S. W. Lee and R. Mittra, "Fourier Transform of a Polygonal Shape Function
and its Application in Electromagnetics," IEEE Trans. Antenna and Propaga.,
Vol. AP-31, pp. 99-103, 1983.
[50] J. H. Richmond, "A Wire-Grid Model for Scattering by Conducting Bodies,"
IEEE Trans. Antenna and Propaga., Vol. Ap-14, p782- ,1966.
[51] J. Mathews and R.L. Walker, Mathematical Methods o f Physics, 2nd Ed.,
W.A. Benjamin, Inc., Menlo Park, California, 1970.
[52] E. F. Knott, "Concept of Cross Section," in Chap.6 of Radar Cross Section,
edited by E. F. Knot e t al, Artech House, Dedham, MA. 1985.
[53] D. Guili, "Polarization Diversity in Radars," Proc. IEEE, Vol. 74, No. 2, pp.
245-269, 1986.
[54] R. C. Johnson, H. A. Ecker, and R. A. Moore, "Compact Range Techniques
and Measurement," IEEE Trans, on Antennas and Propaga., Vol. AP-17, pp.
568-576, Sept., 1969.
[55] M. T. Tuley, "Compact Range Measurement," in Chap. 9 of Techniques of
Radar Reflectivity Measurement, edited by Nicholas C. Currie, Artech House,
Inc., Dedham, MA, 1984.
[56] J. A. Stratton, Electromagnetic Theory, McGraw-Hill Book Co. New York, pp.
480-490, 1940.
[57] Y. W. Kiang, "RCS Mathematical Model," Annual Report, EM Lab, National
Taiwan University, Taipei, R. O. C., 1985
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-239-
[58] H. Stark, J. W. Woods, I. Paul, and R. Hingorani, "An Investigation of Compu­
terized Tomography by Direct Fourier Inversion and Optimum Interpolation,"
IEEE. Trans. Biomedical Engineering, Vol. BME-28, pp.496-505, 1981.
[59] N. N. Bojarski, "A Survey of the Physical Optics Inverse Scattering Identity,"
IEEE Trans. Antenna and Propaga., Vol. AP-30, pp.930-989, 1982.
[60] C. L. Wemer, "Three-Dimensional Imaging of Coherent and Incoherent
Sources Utilizing Wavevector Diversity," Ph.D. dissertation, University of
Pennsylvania, 1987.
[61] D. A. Ausherman, Adam Kozma, J. L. Walker, H. M. Jones, and E. C. Poggio,
"Develoment in Radar Imaging," IEEE Trans, on Aerospace and Electronic
Systems, Vol. AES-20, No.4, pp.363-400, 1984.
[62] R. W. Gerchberg, "Super-Resolution Through Error Energy Reduction," Optica
Acta, Vol. 21, No. 9, pp. 708-720, 1974.
[63] Y. Yamakoshi and T. Sato, "Iterative Image Restoration From Data Available
in Multiple Restricted Regions," Applied Optics, Vol. 21, No. 24, pp.44734480, 1982.
[64] C. O. Lan, K. K. Xu and Glen Wade, "Limited Angle Diffraction Tomography
and Its Application to Planar Scanning Systems," IEEE Trans, on Sonics and
Ultrasonics, Vol. SU-32, No. 1, pp. 9-16, 1985.
[65] S. B. Bowling and S. Lai, "The Use of Linear Prediction for the Interpolation
and Extrapolation of Missing Data and Data Gaps Prior to Spectral Analysis,"
Proc. o f the RADC Spectrum Estimation Workshop, Oct. 3, 4, and 5, 1979, pp
39-49.
[66] S. M. Kay and S. L. Marple, "Spectrum Analysis - A Modem Perspective,"
Proc. o f the IEEE, Vol. 69, pp.1380-1419, November 1981.
[67] J. Capon, "High-Resolution Frequency-Wavenumber Spectrum Analysis," Proc.
IEEE., Vol. 57, pp. 1408-1418, Aug. 1969.
[68] V. F. Pisarenko, "The Retrieval of Harmonics from a Covariance Function,"
Geophysical J. Royal Astronomical Soc., Vol. 33, pp. 347-366,1973
[69] M. L. Van Blaricum and R. Mittra, "Problems and Solutions Associated with
Prony’s Method for Processing Transient Data," IEEE Trans. Antennas and
Propagat., Vo. Ap-26, pp. 174-182, Jan., 1978.
[70] Ralph O. Schmidt, "Multiple Emitter Location and Signal Parameter Estima­
tion," Proc. RADC Spectrum Estimation Workshop Oct. 1979, pp 243-258
[71] R. E. Dubroff, "The Effective Autocorrelation Function of Maximum Entropy
Spectra," Proc. IEEE., Vol. 63, pp.1622-1623, 1975.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-240-
[72] J. Makhoul, "Linear Prediction: A Tutorial Review," Proc. IEEE,
pp.561-580, Apr., 1975.
Vol. 63,
[73] J. P. Burg, "Maximum Entropy Spectral Analysis," in 37th Meeting Society of
Extrapolation Geophysicist (Oklahoma City, OK ), Oct. 31, 1967.
[74] S. Haykin, ed., Nonlinear Methods o f Spectral Analysis, 2nd edition, SpringVerlag, Berlin, 1983.
[75] J. P Burg, "A New Analysis Technique for Time Series Data", NATO
Advanced Study Institute on Signal Processing with Emphasis on Underwater
Acoustics, pp. 12-23, August 1968.
[76] M. I. Skolnik, Introduction to Radar Systems, 2nd edition, McGraw-Hill Book
Company, New York, 1980
[77] T. B. A. Senior, "Backscattering from Resistive Strips," IEEE Trans, on
Antenna and Propaga. Vol. AP-27, pp.808-813, 1979.
[78] O. M. Bucci, G. Franceschetti, "Electromagnetic Scattering by a Half Plane
with Two Faces Impedance," Radio Science, Vol. 11. N o.l, pp. 49-59, 1976.
[79] R. Tiberio, F. Bessi, G. Manara, and G. Pelosi, "Scattering by a Strip with Two
Face Impedances at Edge-on Incidence,” Radio Science, Vol. 17, No. 5,
pp.1199-1210,1982.
[80] A. K. Bhattacharyya, and S. K. Tandon, "Radar Cross Section of a Finite
Planar Structure Coated with a Lossy Dielectric," IEEE Trans, on Antenna and
Propaga. Vol. AP-32, pp. 1003,1007, 1984.
[81] W. H. Emerson, "Electromagnetic Wave Absorbers, Useful Tools for
Engineer," Emerson & Cumming Co.
[82] E. F. Knott, "Radar Cross Section Reduction Using Cylindrical Segments,"
IEEE Trans, on Antenna and Propaga. Vol. AP-24, pp.882-994, 1976.
[83] E. F. Knott, "The thickness Criterion for Single-Layer Radar Absorbents,"
IEEE Trans, on Antenna and Propaga. Vol. AP-27, p.698-701, 1979.
[84] M. C. Lin and Y. W. Kiang, "Natural Frequencies of Dielectric Coated Wires,"
1986 Natioanl Radio Science Meeting, Philadelphia, PA., June 8-13, 1986,
pp.224
[85] J. A. Landt, E. K. Miller, and F. J. Deadrick, "Time Domain Modeling of Non­
linear Loads," IEEE Trans, on Antenna and Propaga., Vol. AP-31, No.l, pp.
121-124, 1983.
[86] R. F. Harrington, J. R. Mautz, "Straight Wires with Arbitrary Excitation and
Loading," IEEE Trans, on Antenna and Propaga. Vol. AP-15, pp. 502-515,
1967.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
-241 -
[87] R. F. Harrington and J. R. Mautz, "Control of Radar Scattering by Reactive
Loading," IEEE Trans, on Antenna and Propaga., Vol. AP-20, pp.446-454,
1972.
[88] R. F. Harrington and J. R. Mautz, ’Pattern Synthesis for Loaded N-port Scatterers,’ IEEE Trans, on Antenna and Propaga., Vol. AP-22, pp. 184-190, 1974.
[89] R. F. Harrington and J. R. Mautz, "Optimization of Radar Cross Section of NPort Loaded Scatterers," IEEE Trans, on Antenna and Propaga., Vol. AP-22,
pp697-701, 1974.
[90] T.K. Liu and F. M. Tesche, "Analysis of Antennas and Scatterers with Non­
linear Loads," IEEE Trans, on Antenna and Propaga., vol. AP-24, No.2, pp.
131-139,1976
[90] J. R. Mautz and R. F. Harrington, "Model Analysis of Loaded N-pc:t Scatter­
ers," IEEE Trans, on Antenna and Propaga., Vol. AP-21, pp. 188-199, 1973.
[91] T. K. Sarker, D. D. Weiner, "Scattering Analysis of Nonlinearly Loaded Anten­
nas," IEEE trans on Antenna and Propaga., Vol. AP-24, No 2, pp.125-131,
1976.
[92] K. M. Chen and V. Liepa, "The Minimization of the Back Scattering of a
Cylinder by Central Loading," IEEE Trans, on Antenna and Propaga., Vol.
AP-12, pp. 576-582, 1964.
[93] I. P. Yu and L. C. Shen, "Modification of Scattered Field of Long Wire by
Multiple Impedance Loading," IEEE Trans, on Antenna and Propaga., Vol.
AP-19, pp. 554-557, 1971.
[94] T.T. Wu and R. W. P. King, "The Cylindrical Antenna with Nonreflecting
Resistive Loading," IEEE Trans. Antennas and Propaga., Vol. AP-13, pp.
369-373, May, 1965.
[95] L. C. Shen, "An Experimental Study of the Antenna with Nonreflecting Resis­
tive Loading." IEEE Trans. Antennas and Propaga., Vol. AP-15, pp. 606-611,
Sept., 1967.'
[96] B. L. J. Rao, J.E. Ferris and W. E. Zimerman, "Broadband Characteristics of
Cylindrical Antennas with Exponentially Tapered Capacitive Loading," IEEE
trans on Antenna and Propaga., Vol. AP-17, pp. 145-151,1969.
[97] R. B. Green, "The General Theory of Antenna Scattering," Antenna Labora­
tory, Ohio State University, Report No. 223-17, November 1963.
[98] H. Schuman, "Time-Domain Scattering from a Nonlinearly Loaded Wire,"
IEEE Trans, on Antenna and Propaga., Vol. AP-22, pp. 611-613, 1974.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
- 242-
[99] A. R. Djordjevic and T. K. Sarker, "Transcient Analysis of Electromagnetic
Systems with Multiple Lumped Nonlinear Loads," IEEE Trans, on Antenna
and Propaga., Vol. AP-33, No.5, pp. 533-539, 1985.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Документ
Категория
Без категории
Просмотров
0
Размер файла
9 733 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа