close

Вход

Забыли?

вход по аккаунту

?

504

код для вставкиСкачать
J Sci Food Agric 79 :86–90 (1999)
Journal of the Science of Food and Agriculture
Maturity determination of fresh dates by near
infrared spectrometry
Ze’ev Schmilovitch,1* Aharon Hoffman,1 Haim Egozi,1 Rachel Ben-Zvi,2
Zvi Berns tein2 and Victor Alchanatis 3
1 Ins titute of Agricultural Engineering , Department of Pos t -Harves t Technologies and Quality Control ARO , Volcani Center , Bet Dagan ,
Is rael
2 Zemach Regional Laboratories , Jordan Valley , Is rael
3 Ins titute of Agricultural Engineering , Department of Tes ting and Ins trumentation , ARO , Volcani Center , Bet Dagan , Is rael
Abstract : In Israel, fresh dates are normally harvested unripe and stored in a controlled environment
until marketed. Proper ripening depends on maturity at harvest time. Sugar and water contents of
the dates are considered important maturity attributes. Decision-making on the harvesting schedule
for each section in a plantation of fresh dates (variety Hayani) is one of the major problems for the
growers. In order to estimate the optimum harvest time, conventional laboratory methods have been
used to assess the maturity of samples of dates ; methods which by nature are destructive, manual and
time consuming. A semi-automatic system for maturity determination of fresh dates has been developed, tested and operated. It combines a near infrared (NIR) spectrometer with a step-wise cell conveyor, both of which are controlled by a PC. The NIR models were based on measuring the water and
total soluble solids (TSS) contents. The calibration models and the prediction results had a standard
error of prediction (SEP) of 1% for TSS and water contents. The correlation coefficient R between
TSS and water contents as predicted by NIR models and as measured in routine laboratory destructive tests was 0.9. The system was capable of testing 100 dates in 3 min.
( 1999 Society of Chemical Industry
Keywords : non-destructive ; quality ; control ; harvest ; automation ; sorting
INTRODUCTION
Decision-making on the harvesting schedule for each
section in a plantation of fresh dates (Phoenix dactylifera L, variety Hayani) is one of the major problems for the growers and it requires investment of
time and labour. The main factor in this decision is
the maturity of the dates. In Israel, fresh dates are
harvested at an unripe stage (coloured red) and
stored in cooled conditions, until marketed. In this
manner, pre-scheduled ripening is prevented, as is
necessary because the shelf-life of fresh dates is very
short after ripening (turning black).
The main factor that inýuences proper future
ripening is sugar content. During the ünal stages of
growing dates accumulate sugars and their water
content falls.1 Fruits with sugar content below a
critical value will not ripen properly and will have
poor colour (brown) and taste. Water content and
sugar content are highly correlated, and the former
has been proven to be a suitable factor for assessing
sugar content in fresh dates.2 Dates with water
content below 66% or with sugar content above
30.5% (in terms of TSS content) are regarded as
mature and will ripen properly.3
Each plantation is divided into several sections, in
each of which the trees are similar in age and agricultural conditions. However, the maturity of the fruits
is distributed and it progresses during the season. In
harvesting fresh dates (‘wet dates’) the harvest of
each section of a plantation is usually single pass,
non-selective work. Therefore, the harvest schedule
needs to be optimised according to the maturity of
the section, in order to provide about 80% mature
fruits. Early harvesting might gather a low level of
mature fruits even below 60%, and so reduce efficiency and proüts. Late harvesting might cause a loss
of ripened fruit, unsuitable for storage. Furthermore,
decisions on allocating equipment and workers
should be taken as early as possible on harvest days,
according to the maturity of the section.
Several destructive methods are in use for determination of maturity of dates. Measuring the Brix of
date juice by optical refractometry provides information on sugar content terms of TSS.1 Water content
can be measured by drying the dates in a vacuum
oven for 48–72 h at a temperature of 65¡C. Soaking a
sample of dates in vinegar solution accelerates ripening of mature fruits, so that maturity can be evalu-
* Corres pondence to : Ze’ev Schmilovitch, Ins titute of Agricultural
Engineering, Department of Pos t-harves t Technologies and
Quality Control ARO, Volcani Center, Bet Dagan, Is rael
(Received 27 January 1998 ; revis ed vers ion received 17 April
1998 ; accepted 4 May 1998 )
( 1999 Society of Chemical Industry. J Sci Food Agric 0022–5142/99/$17.50
86
Maturity determination of fresh dates by NIR spectrometry
ated by counting the percentage of black dates after
12 h (overnight). All the above methods are either
time or labour consuming.
A preliminary study on fresh dates2 proved the
feasibility of using NIR analysis for determination of
water content and TSS. This work was conducted
mainly with fresh dates before sorting : they had been
in cooled storage prior to the test.
Figure 1. Schematic view of the firs t s ys tem.
OBJECTIVES
The objective of this project was to develop a rapid,
non-destructive and reliable method for determination of maturity of fresh dates in the plantation,
using NIR analysis. The requirements from such a
system were deüned as follows :
(1)
(2)
(3)
(4)
determination of water and TSS contents ;
semi-automatic feeding with simple operation ;
testing each date in 2 s ;
low maintenance demands.
METHODS AND MATERIALS
Three stages of development are described in this
paper. Each stage covers a year of development of a
system intended to be tested and operated during the
harvesting season. In this project, a commercial NIR
scanner (Quantum 1200, manufactured by LTI,
MD, USA) was used in reýectance mode, in the
range of 1200–2400 nm, using a über-optic bundle in
the sensing compartment.
In the ürst season (1994), a set of 120 dates was
collected from each of six plantations, in order to
create a representive sample of the date population.
Each date was scanned and its NIR spectrum was
collected. Each date was divided to two halves, one
of which was used for measuring water content by
drying the date in a vacuum oven for 48 h at a temperature of 65¡C. The other half was used to determine the total soluble solids (TSS) content by
measuring the Brix of extracted juice by means of
optical refractometry (ATAGO, model PR101,
J apan).
The data from all plantations were combined and
divided randomly so that 300 dates provided the calibration database and the remainder were used for
evaluating the model prediction error. Processed
spectra such as the ürst derivative, log(1/R), and its
ürst and second derivatives were analysed by principal component analysis, partial least squares (PLS)
and Multi-Linear regressions by using the Spectra
Metrix and Light Cal software packages (LTI, MD,
USA). The desired NIR model should have a low
error of prediction, with as few factors as possible,
and observation of this criterion led to a calibration
procedure which used a model based on the ürst
derivative of the spectra and four-factor PLS regression. The PLS regression models presented in this
paper relate the ürst derivative NIR spectra to the
J Sci Food Agric 79 :86–90 (1999)
water and TSS contents in each tested date. The
same underlying model (four-factor PLS, of the ürst
derivative of the spectra) was also used in the two
subsequent systems with new calibration sets in each
season.
DEVELOPMENT OF THE SYSTEM
First system
The system that was developed in the ürst stage
(harvest season of 1994) is schematically presented in
Fig 1. This system served to prove the ability to
work in automatic mode with a single cell. A pneumatic cylinder moved the cell, containing a single
date into a position such that the holes in the cell and
in the device were aligned with the über optic bundle
and NIR detectors. The date was covered and was
held in position by a soft rubber wheel. In that position the date was scanned by the NIR system. After
scanning, the cell was pushed out for the date to be
replaced with another, and when the whole set had
been scanned the results were analyzed and output to
the screen. The ürst system formed the basis for the
development of a more convenient continuousoperation system.
Second system
For the harvest season of 1995 (September) a new
system was developed. A schematic view of this
device is presented in Fig 2. A step-wise cell conveyor was designed and built, such that the system used
the same optical conditions and sensor conüguration
as the ürst system. The optical conüguration of the
sensing area is shown in Fig 3. A PC, as in the ürst
stage, controlled both the conveyor and the spectrometer. A calibration procedure was conducted, in
Figure 2. Schematic view of the s econd s ys tem.
87
Z Schmilovitch et al
Figure 4. Optical configuration of the s ens or for the third s ys tem.
Figure 3. Optical configuration of the s ens or for the s econd
s ys tem.
order to update the NIR models, as with the previous system.
The system was installed in the Zemach-Tmarim
packinghouse. At the beginning of the season, NIR
tests were conducted in parallel with vinegar solution
soaking.
The standard deviation of the 1994 season samples
had been 7.7 and 3.3% for water and TSS contents
respectively. On the basis of these ügures, the selection of a sample set size of 100 dates ensured that the
conüdence intervals of the mean was 1.5 and 0.4%
for water and TSS contents respectively, which
meets the requirements of the packinghouse.
In order to evaluate the system performance, a
comparison experiment was conducted. Eleven
hundred dates were mixed and divided into 11 sets.
Six sets were soaked in vinegar solution and üve
were transferred to cool storage for natural ripening.
The result of this test is presented in Table 1. Ripening percentages of the vinegar and the storage sets
Table 1. Comparis on of res ults of NIR, vinegar and s torage tes ts
Sample number
Ripening percentage
Vinegar
tes t
1
2
3
4
5
6
Mean
SD
7
8
9
10
11
Mean
SD
88
41
51
43
38
46
43.80
4.97
70
61
68
67
66
66.40
3.36
TSS model
NIR
NIR water
content model
49
41
51
46
40
43
45.00
4.43
49
46
59
45
46
49.00
5.79
48
41
49
44
41
43
44.33
3.44
49
46
56
45
47
48.60
4.39
were evaluated according to the number of dates that
turned black in each set. Ripening percentages were
predicted with the NIR models, according to the
number of dates with TSS content higher than
30.5% or with water content less than 66%.3
Third system
Although the second system had worked quite well,
some improvements were considered necessary for
the 1996 season. These improvements yielded a completely new device. Although there was no contact
between the scanned fruit and the über-optic bundle
or the NIR detector, in contrast to other NIR
devices, as described by other researchers4h6 and
manufacturers (Fantec, Shizuka, J apan), maintaining
the same optical condition (Fig 3) caused a frictional
problem between moving parts. To overcome this,
an air purging system had to be installed and daily
cleaning was recommended. To avoid the additional
component of the system, a new conüguration was
developed, which is schematically described in Fig 4.
The sufficient signal-to-noise ratio that was achieved
in the experiments conducted with this conüguration
led to a new design for the stepwise conveyor ; it is
described schematically in Fig 5.
User-friendly and more efficient software was
developed for operation of the machine. This software provided overall control of the machine, the
menu-driven user-machine interface, and the automatic updating of the results database. Using this
software, the operator registers the identity of the set
being tested. The results obtained from the tested
Figure 5. Schematic view of the third s ys tem.
J Sci Food Agric 79 :86–90 (1999)
Maturity determination of fresh dates by NIR spectrometry
Figure 6. The third s ys tem as
ins talled in the ZemachTmarim packing hous e.
set, with its identiücation, are printed out as a hard
copy by an attached printer and recorded in an
archive üle. The system was installed and operated
in the Zemach-Tmarim packinghouse (Fig 6).
RESULTS
The results of prediction with NIR models (using
the validation set) during the operation of the ürst
system are presented in Figs 7 and 8. The prediction
results had a Standard Error of Prediction (SEP) of
1% for both TSS and water contents. The correlation coefficient, R between the predictions of TSS
and water contents based on NIR models and on
routine destructive laboratory tests was 0.9. Similar
prediction results were obtained for validation sets
for the second and the third systems.
Cumulative distribution curves of water and TSS
contents, as calculated from the destructive tests
(refractometer and oven) and predicted by NIR
models, are presented in Figs 9 and 10. The small
distance between the two curves in each ügure indicates close similarity between the sample distributions. Hence, the water and TSS content distribution
Figure 9. Comparis on of the s ample dis tribution bas ed on the
oven res ults with the dis tribution bas ed on NIR model res ults .
Figure 7. Prediction res ults by the NIR s ys tem vers us laboratory
meas urements of water content of Hayani dates .
Figure 8. Prediction res ults by the NIR s ys tem vers us laboratory
meas urements of TSS content of Hayani dates .
J Sci Food Agric 79 :86–90 (1999)
Figure 10. Comparis on of the s ample dis tribution bas ed on the
refractometer meas urements with the dis tribution bas ed on NIR
model res ults .
89
Z Schmilovitch et al
results showed similarity between the two methods
in the average and standard deviation values. Furthermore, examination of the curve showed that the
proportion of mature dates, ie dates whose water
content is below 66% or sugar content is above
30.5%, was similarly predicted by the diþerent
methods (0.38–0.40).
During the second season it was established that
the information provided to the growers should be
based on three limits :
(A) the percentage of the dates in the set that have
a water content below 66% ;
(B) the percentage of the dates in the set that have
a water content below 67% ;
(C) the percentage of the dates in the set that have
a water content below 68%.
These limits furnished the grower with the information needed for the subsequent 5 days, since limit A
represents the dates that are already mature, while B
and C represent the predictions of mature dates after
2 and 5 days, respectively. The slope that could be
derived from those three values represents the homogeneity of the sampled section ; a steep slope represents a uniform section while a moderated slope
indicates a widely spread distribution.
The comparison between the six sets which were
soaked in vinegar solution and the üve sets that had
been transferred to cool storage for natural ripening
is presented in Table 1. Comparison between the
standard deviations of the samples shown in the
upper part of Table 1, shows that the consistency of
the NIR water content model was better than that of
the vinegar test. The higher post-ripening percentages, shown in the lower part of Table 1, are related
to the fact that after one month in storage, some of
the dates with low sugar contents ripened but with a
poor taste.
During the third year, the new system was calibrated, and it operated with the same success.
Growers were furnished with the same information
as with the second system. The measuring rate of the
third system was 2 s per date. Although a sample of
100 dates was measured in 3 min, the working time
needed to complete a full test procedure was 5 min.
The number of sets presented for testing was 50 per
day at the most, requiring a maximum 4 h of operation per day. Thus, the operator was also able to
carry out acceptance inspections for those growers
who had already started to harvest and supply dates
to the packinghouse.
CONCLUSIONS
Destructive tests for maturity determination of
Hayani dates were not used by the packinghouse
90
during the 1996 harvest season. The NIR system was
accepted by the packinghouse experts and by the
growers as the tool for harvesting schedule decisionmaking. Acceptance inspections were conducted on
incoming shipments in order to ensure that decisions
had been made correctly ; if necessary some supplies
were held. Payment according to acceptance inspection remains a subject for future discussions between
the growers and the production management of the
warehouse.
It is possible to increase the rate of the system to
one date per second by using more dedicated software. Using more advanced (and currently more
expensive) spectrometers might increase the rate to
the limits of mechanical components (about four
dates per second per channel). The NIR method and
the system described in this paper might be considered as a basis for a sorting system for harvested
fresh dates in the future, depending on reduction in
prices of the components and increasing rate of scanning.
ACKNOWLEDGEMENT
Contribution from the Agricultural Research Organization, The Volcani Center, Bet Dagan, Israel, No
2001-E, 1996 series.
REFERENCES
1 Dowson VH and Aten WA, Dates Handling, Processing and
Packing. FAO of the UN, Rome, Italy (1962).
2 Schmilovitch Z, Bernstein Z, Austerweill M, Zaltzman A and
Dull BG, Fresh date sorting by NIRS, in Making Light
Work : Advances in Near Infrared Spectroscopy. Ed by Murry
I and Cow IA. Ian Michael Publications, UK, pp 425–429
(1992).
3 Stoller S, Growing Dates in Israel. Hakibbutz Hameuchad Publication, Israel, p 213 (in Hebrew) (1989).
4 Kawano S, Satto T and Iwamoto M, Determination of sugars in
satsuma orange using NIR transmittance, in Making Light
Work : Advances in Near Infrared Spectroscopy, Ed by Murry
I and Cow IA, Ian Michael Publications, UK, pp 387–395
(1992).
5 Kawano S, Abe H and Iwamoto M, Novel applications of nondestructive techniques for quality evaluation of fruits and
vegetables in J apan, in Nondestructive Techniques for Quality
Evaluation of Fruits and Vegetables (Proceeding in the International Workshop funded by BARD, Spokane, WA, USA).
ASAE publication 05-94, pp 1–7 (1993).
Light-Cal Plus, Software, Rockville, LT Industries Inc, MD,
USA (1989).
6 Slaughter DC, Nondestructive determination of internal quality
in peaches and nectarines. Trans ASAE 38(2):617–623
(1995).
Spectra-Metrix 1.8, Software, LT Industries Inc, Rockville,
MD, USA (1989).
J Sci Food Agric 79 :86–90 (1999)
Документ
Категория
Без категории
Просмотров
7
Размер файла
194 Кб
Теги
504
1/--страниц
Пожаловаться на содержимое документа