1 - Color y Humedad de Soya Deshidratada Con Imagenes Hiperespectrales

8/13/2019 1 - Color y Humedad de Soya Deshidratada Con Imagenes Hiperespectrales

1/7

Journal of Food Engineering 128 (2014) 24 30

Prediction of color and moisture content for vegetable soybean duringdrying using hyperspectral imaging technology

Min Huang a, b, Qingguo Wang a, Min Zhang b,, Qibing Zhu a a Key Laboratory of Advanced Process Control for Light Industry (Ministry of Education), Jiangnan University, 214122 Wuxi, Jiangsu, China b State KeyLaboratory of Food Science and Technology, Jiangnan University, 214122 Wuxi, Jiangsu, China

a r t i c l ei n f o

Article history: Received 8 July 2013 Received in revised form 2 December 2013 Accepted 5 December 2013 Availableonline 12 December 2013

Keywords: Dried soybean Color Image entropy Hyperspectral imaging Moisture content Prediction modeling

a b s t r a c t

Dried soybean is among the most popular snack foods consumed in numerous countries, and its quality has receivedconsiderable attention from processors and consumers. Color and moisture content are two critical parameters used to evaluatedried soybean quality. This study thus aimed to develop regression models for predicting the color and moisture content ofsoybeans simultaneously during the drying process using a hyperspectral imaging technique. Hyperspectral reflectance imageswere acquired from fresh and dried soybeans over the spectral region between 400 and 1000 nm for 270 samples. After theautomatic segmentation of soybean images at each wavelength based on an active contour model, mean reflectance and imageentropy parameters were extracted and tested separately and in combination for predicting the color and moisture content ofthe processed soybeans. Predicting models were built using the partial least squares regression method. Better predictionresults for both color and moisture content were achieved using the mean reflectance data (with correlation coefficients or R P =0.862 and root-mean-square errors of prediction or RMSEP = 1.04 for color, as well as R P = 0.971 and RMSEP = 4.7 % formoisture content) than when using entropy data (R P = 0.839 and RMSEP = 1.14 for color, as well as R P = 0.901 and RMSEP =9.2% for moisture content). However, the integration of mean reflectance and entropy data did not show significantimprovements in predicting the color or moisture content. Overall, a simple hyperspectral imaging technique involving rapidimage preprocessing and single spectral features showed significant potential in measuring the color and moisture content ofsoybeans simultaneously during the drying process.

2013 Elsevier Ltd. All rights reserved.

1. Introduction

Vegetable soybean [Glycine Max (L.) Merrill], also known by theJapanese term edamame, is a so ybean harvested at approximately 80%maturity ( Hu et al., 2006b ). Soybean is popularly consumed after blanching inChina, Korea, Japan, and other countries for its rich protein, fat, calcium,

vitamin, and diet fiber content. Soybean also has potential for cancer prevention and suppression owing to its high genistein content ( Hu et al.,2006b, 2007; Hou et al., 2011 ).

Color and moisture content are two of the most important parameters inevaluating the drying quality of dried soybeans ( Hu et al., 2006a ). Colormeasurements of vegetable soybeans are performed using conventionalcolorimeters and spectrophotometers after drying. However, these traditionalinstrumental techniques are time-consuming because of the repeatedmeasurements required to obtain a representative color profile and to reducethe measurement error ( Hu et al., 2006b, 2007 ). Moreover, these instruments

are designed for color measurements on flat surfaces rather than on curvedsurfaces, which are found in soybeans. The uncertainty of these instrumentalmeasurements might introduce further error in analysis ( Aguilera, 2003 ).

The gravimetric oven method and Karl Fisher titration are commonly usedlaboratory methods for moisture content measurements of agricultural foodsand their products. These methods are destructive measurements, such that the

same samples cannot be used for furtheranalysis. Moreover, current methods for measuring color and moisture contentcannot measure the two parameters simultaneously. Moreover, existingmethods are only suitable for testing a small number of vegetable soybeans.

Rapid nondestructive technologies for measuring the drying qualities ofagricultural products and food products have been studied extensively. Among

these approaches, machine vision and near-infrared spectroscopy are the twomain methods ( Fernndez et al., 2005; Mendoza et al., 2006; Faustino et al.,2007; Lucas et al., 2008; Wu et al., 2010; Romano et al., 2012 ). However, theconventional machine vision method can only acquire average imageinformation within the visible range (i.e., external characteris-

Corresponding author. Fax: +86 510 580 7976. E-mail address: [email protected] (M. Zhang).

0260-8774/$ - see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008 tics of the sample from grayscale or color images). Near-infraredspectroscopy, although used in a wide range of wavelengths, can

only acquire spectral information and cannot obtain the spatialinformation of the samples. Thus, these methods have limitationsin that they cannot provide spectral and spatial information

simultaneously, which may result in the loss of useful information.As a relatively novel non-destructive technology, hyperspectral imaging

integrates the advantages of machine vision and visibleinfraredspectroscopy, while overcoming the drawbacks of both techniques whenused alone. Hyperspectral imaging can provide more detailed or completeinformation, including internal structure characteristics, morphologicalinformation, and chemical composition ( Huang et al., 2013 ). Thistechnology has been applied to the nondestructive measurement of

agricultural products for evaluating internal quality ( Liu et al., 2006; Arianaand Lu, 2008 ; Huang et al., 2010; Li et al., 2012 ) and pesticide residues ( DelFiore et al., 2010; Shahin and Symons, 2011; Peng et al., 2011 ). Thus, this

Contents lists available at ScienceDirec t

Journal of Food Engineering

ournal homepage: www.elsev i er.com/locate/jfooden g
http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008http://www.sciencedirect.com/science/journal/02608774http://www.sciencedirect.com/science/journal/02608774http://www.sciencedirect.com/science/journal/02608774http://www.sciencedirect.com/science/journal/02608774http://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.elsevier.com/locate/jfoodenghttp://www.sciencedirect.com/science/journal/02608774http://www.sciencedirect.com/science/journal/02608774http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008http://dx.doi.org/10.1016/j.jfoodeng.2013.12.008


2/7

M. Huang et al. /Journal of Food En gineering 128 (2014) 24 30 25

technology may also be used as an alternative for predicting thecolor and moisture content of vegetable soybeans during drying.

The overall objective of this study is to use a hyperspectralreflectance imaging technique in the wavelength range of 400 1000 nm for predicting the color and moisture content ofvegetable soybeans simultaneously during drying. The specificobjectives are as follows:

To extract image traits from preprocessed hyperspectralreflectance images of dried soybeans using mean and entropymethods; and

To evaluate the capability of partial least squares regression(PLSR) models for predicting the color and moisture contentof vegetable soybeans during drying.

2. Materials and methods

2.1. Raw materials

Two hundred and seventy fresh soybeans [Glycine Max (L.)Merrill] harvested from the Garden of Haitong Food Company inCixi, Zhejiang Province, during the 2012 harvest season wereused in this study. The soybeans were washed, peeled and

blanched using the microwave heating method and then stored at4 C and 95 % relative humidity in a refrigerator before theexperiments. The soybeans were used within 3 days.

2.2. Microwave-assisted pulse-spouted bed vacuum-drying

Vegetable soybeans were dried using a high-precisionlaboratory dryer developed at the State Key Laboratory of FoodScience and Technology, Jiangnan University, China. Thismicrowave-assisted pulse-spouted bed vacuum-drying (PSMVD)experimental system essentially consisted of seven units: (a) acylindrical multimode microwave cavity, (b) two circular vacuumdrying chambers, (c) a pulse-spouted system as a nitrogen gassource, (d) a refrigeration system with a set of air-coolingrefrigeration compressor units, (e) a vacuum system, (f) twoenergy supply systems, and (g) a water load system to preventmagnetron from overheating using a cooling/heating circulatingwater unit. A detailed description of the dryer system is given byWang et al. (2013) .

In this study, the experimental parameters were set as follows:(1) the pressure was set at 9 1 kPa; (2) the power was set to 516W; and (3) the samples were spouted in the preselected timeinterval of 1 s and held for 3 s by allowing nitrogen gas to flow

periodically into the drying chamber. Fresh soybeans with a massof 200 g were used for each trial. To achieve broad sampledistribution of color and moisture content, eight groups atdifferent drying times (from 10 min to 80 min, in steps of 10 min)were tested. The experiments were replicated thrice for eachdrying condition. Fresh and dried samples were measured usingthe hyperspectral imaging system and then tested using referencemethods for color and moisture content.

2.3. Hyperspectral reflectance image acquisition

An in-house developed line-scan hyperspectral reflectanceimaging system was used to acquire hyperspectral reflectance imagesof soybeans. The hyperspectral reflectance image system mainlyconsisted of a hyperspectral imaging unit, a light source, and asample handling platform. The hyperspectral imaging unit compriseda back-illuminated 1392 1024 pixel charge-coupled device (CCD)camera (Pixelfly QE IC285AL, Cooke, USA), an imaging

spectrograph (1003A- 10140 Hyperspc VNIR C -Series, Headwall

Photonics Inc., Fitchburg, USA) with a 25 lm slit covering an effective range of400 1000 nm and connected to a zoom lens (10004A-21226 Lens, F/1.4 FL23mm, Standard Barrel, C-Mount., USA), and a computer for controlling thecamera and acquiring the images. The light source system consisted of a 150 WDC light source (halogen lamp, EKE, 3250K, Techniquip, USA) and a singleoptic fiber coupled to a 9 inch parallel light lamp. The sample handling unitconsisted of a horizontal motorized stage. Ten soybeans were placed onto a 20cm 20 cm black background board in two rows and perpendicular to thescanning line of the hyperspectral imaging unit (see Fig. 1 ).

For each group of samples, 625 scans covering a 50 mm distance wereacquired at an exposure time of 250 ms for each hyperspectral reflectanceimage. The hyperspectral imaging system had 0.15 mm/pixel spatial resolutionand 0.644 nm/pixel spectral resolution covering the spectral region of 400 1000nm using a 1392 pixel camera. After 10 spectral binning operations, theresultant hyperspectral reflectance images had 6.44 nm/pixel spectral resolutionand 94 wavelengths. Thus, a spatial block of a 1392 625 94 image was created,which was represented by a 2-D image with x-axis and y-axis coordinateinformation. Another axis was represented by spectral information. Thedarkness and reflectance images of white Teflon were also acquired for everysix groups of samples and used as reference to obtain relative reflectanceimages.

2.4. Reference measurements

A CR-400 Chroma Meter (Konica Minolta Sensing, Inc., Japan) was usedfor soybean surface color measurements. Color difference (DE) was used todescribe the color change in the fresh and dried samples and was calculated asfollows:

Fig. 1. Schematic of the hyperspectral reflectance imaging system. DE qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffif fiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi L0 L2

a0 a2 b0 b2 1

where L 0; a0 and b 0 are the color readings of standard white plate at D 65illumination (L 0 97:06; a0 0:04; and b 0 2:01). Two measurements were

performed per soybean (at each side of the grain), and the average value wasrecorded as the reference color of each grain sample.

The moisture content, expressed in percent wet basis (% w.b.), wasmeasured by the gravimetric method using a convection oven (GB/T8858-88,

National Standard of China). The samples dried at different times using thePSMVD experimental system were placed into the oven (Binder FED, Berlin,Germany) at 105 C for 7 8 h until they reached a constant weight. The weightwas measured using an analytical balance (Hengping FA1104, Shanghai, China;0.0001 g). Finally, moisture content was calculated.


3/7

26 M. Huang et al./ Journal of Food Engineering 128 (2014) 24 30

2.5. Data analysis

2.5.1. Image preprocessingTo compensate for the light source variation effect, relative

reflectance was used and obtained using the following equation:

TR T S TD 2

TF TD where T R is the relative reflectance; T S and T F are the intensity of thesample and of the reference (i.e., Teflon), respectively; and T D is thedark signal for the CCD detector. Thus, all further analyses wereconducted on the relative reflectance images.

Segmentation of the soybeans from the hyperspectral image background is a key step in extracting the image features used todevelop the prediction models. Among the image segmentationmethods, the global threshold approach is a traditional technique withthe advantages of easy calculation and high efficiency. However, thesegmented image is entirely dependent on the selected thresholdvalue. The active contour model (ACM) is taken as a suitablesolution to mitigate the problem of threshold selection. The main ideaof ACM is to define the energy function of an arbitrary initial closedcurve as well as to drive the evolution of the closed curve byminimizing the energy function until the closed curve reaches theobject boundaries. ACM has more advantages for imagesegmentation because it is region-based, thus enabling the delineationof regions defined by smooth intensity. Moreover, this method doesnot impose any significant initialization constraint ( Chan and Vese,2001 ). The energy function E(C, c 1, c2) is given by

Z EC;c1;c2 lLC vS1C k 1 jI c1 j2dxdy k 2

C Z inside

jI c2 j2dxdy 3 Coutside

where C is the variable curve. R and K(x, y) = exp ( kx yk2/s2) ,

which depend on C, are the average intensities of the target regionand the background region, respectively. lP 0 ;vP 0, C, and C arefixed parameters. L(C) and S 1(C) are respectively the length of theevolving curve C and the area of the region inside C and areexpressed as follows: Z

LC d/x;yjr/x;yjdxdy Z X

S1C H/x;ydxdy 4 X where s is the definition domain of the original image C;

and s and C are the Heaviside and Dirac functions, respectively. Byconstructing an Euler Lagrange equation for /, the / function isupdated by recalculating c 1 and c 2 in each process of curve evolution.Continuous iterations can be attained once the evolution curvereaches the ultimate target boundary to yield the final segmentation

results. Fig. 2 shows the contour segmentation result of representativefresh and dried soybeans (after 30 and 60 min) at a 718.2 nmwavelength.

2.5.2. Feature extractionAfter the automatic segmentation of the image background at

each wavelength, a large amount of spatial and spectralinformation was obtained from the true image of each soybean.Image analysis mainly aims at the effective extraction of usefulinformation. Considering the apparent changes in the soybeansurface and the chemical composition of soybean during thedrying process, mean reflectance and entropy methods wereapplied to extract and predict the color and water contentcharacteristics of the dried grains. Similar to near-infraredspectrum, the mean value of relative reflectance images (mean ofthe i pixel intensity in the image at each wavelength) provides

physical and chemical information on the samples. Image entropy can beused to describe the uniformity or randomness of the pixel intensity in thereflectance image ( Zhu et al., 2013; Wu et al., 2008 ) that resulted from thetextural changes during drying of the samples. The mean reflectance andentropy are expressed in following equations:

1

XM X N TR i; j 5

R M N i1 j1

Fig. 2. Contour segmentation results for fresh and dried soybeans at a wavelength of 718.2 nm. XM X N

H pijlog2 pij i1 j1

XM X N pij TR i; j= TR i; j

i1 j1

where R is mean reflectance; H is entropy; T R (i, j) is the relative reflectanceintensity of pixel (i, j), (i = 1, 2, ...,M j = 1, 2, ...,N); and M, N are thenumber of total horizontal and vertical pixels for a soybean, respectively.

2.5.3. Regression analysisPLSR was used to predict the color and moisture content using the mean

reflectance and entropy feature parameters. In developing the calibrationmodel, 270 samples were randomly divided into two sets: 3/4 of the samples

( a Contour segmentation results for fresh soybeans

( b ) Contour segmentation results for dried soybeans (30 min)

( c ) Contour segmentation results for dried soybeans (60 min)


4/7


were used for calibration, and the remaining 1/4 was used for prediction (independent validation) after spectral pre-processingusing the five-point moving average smooth method. For themean reflectance and entropy spectra, the mean-center algorithmwas used prior to analysis, whereas the autoscale algorithm wasemployed for mean reflectance combined entropy spectra becausetheir variables have different units. The PLSR models (i.e.,selection of the appropriate number of latent variables) weredetermined by a full cross-validation of the calibration samplesusing leave-one-out cross validation until the rootmean-squareerror of cross validation (RMSECV) reached the minimum. Aftera calibration model was developed, this model was used to predictthe independent set of samples that had not been used in the

calibration. The calibration and prediction results may varydepending on how the calibration and prediction samples wereactually selected. To compare the performance of the meanreflectance and entropy methods better for predicting the colorand moisture content of dried soybeans, the calibration andvalidation procedure described above was repeated 10 times byelecting a random set of samples. The average values of thecorrelation coefficient (i.e., R C, R CV, and R P) and root-mean-square error for calibration and prediction (i.e., RMSEC,RMSECV, and RMSEP) were calculated to evaluate the

performance of the models. The PLSR was run in Matlab (2009b)with PLS-Toolbox 5.0 ( Eigenvector

Research, Inc., Wenatchee, WA, USA). Finally, a t-test was performed to determine the statistical differences between meanreflectance and entropy methods for predicting the two evaluatedquality parameters.

3. Results

The instrumental measurement of color difference (DE) for the270 samples, including fresh and dried soybeans of different drying

times (from 10 min to 80 min), ranged from 49.9 to 63.2. The mean

value of color difference was 54.8, with a standard deviation of 2.1. Themoisture content distribution was in the range of 4.9 67.7 %, with a mean valueof 29.0% and a standard deviation of 20.2 %.

3.1. RGB images of fresh and dried soybeans

Representative color images of test samples, including fresh and dried

soybeans (at 10, 30, 50, and 70 min) with apparent differences in moisturecontent, are shown in Fig. 3 . The samples with higher moisture content appearsmooth and rounded with a homogenous glossy surface. As drying timeincreases, the moisture content decreases, thus causing the samples to havetextured and more brittle surfaces. Shrinkage also occurs as the soybean

samples lose moisture content. At longer periods of drying (i.e., longer than 50min as shown in this experiment), the soybean surfaces show some cracks and

brittle-appearing fractures.

Wavelength (nm)

Fig. 3. Color images of fresh and dried soybeans at different drying times (10, 30, 50, and 70 min). (For interpretation of the references to color in this figure legend, the reader is referred to the webersion of this article.)

(54.7 % moisture content )

( d) Soybeans dried for 50 min(13.4 % moisture content )

( e) Soybeans dried for 70 min

(30.4 % moisture content)

(9.3% moisture content )

( a ) Fresh soybeans ( 66.4% moisture content)

( b ) Soybeans dried for 10 min ( c) Soybeans dried for 30 min


5/7


Wavelength (nm)

Fig. 4. Relative reflectance (a) and entropy (b) for a fresh soybean at different dryingtimes (10, 30, 50, and 70 min).

3.2. Characterization of relative reflectance and entropy spectra

Fig. 4 (a) shows the representative relative reflectance spectrafor fresh and processed soybeans at different drying times (10, 30,50, and 70 min). Typical downward peaks were observed in thevisible range around 435, 535, and 660 nm, corresponding tocarotenoid, anthocyanin, and chlorophyll absorption, respectively.Over the wavelength region of 400 920 nm, the relativereflectance for fresh soybeans was generally greater than that forthe dried samples. Over the same wavelength range, an evident de-

Table 1 Average of 10 calibration and prediction results for color by PLSR models forsoybeans. crease in relative intensity was observed at the initial drying

periods between 10 and 30 min. Thereafter, the intensity ofreflected light tended to increase as observed at 50 and 70 min ofdrying.

At the initial periods of drying, microstructural changesoccurred on the surface of the grains because of the evaporation ofwater in the immediate surroundings of the grain surface. Duringthis initial period, vapor diffusion is the predominant mechanism,and the rate of evaporation or the rate of drying remains constant,thus resulting in discolorations and textural changes on the grainsurface going from slight to moderate, as observed between 10

and 30 min of drying ( Fig. 3 (b) and (c)). However, as soon as the

outer layers of the grain cells on the surface become unsaturated withmoisture, the drying rate falls sharply because the diameters of pores andcapillaries decrease. This condition results in the shrinkage and compactnessof the surface microstructure. This compact and brittle structure of thesoybean probably explains the increase in relative intensity values observedat 50 and 70 min of drying ( Fig. 4 (a)).

Fig. 4 (b) shows the representative spectra of entropy for fresh and processed soybeans at different drying times (10, 30, 50, and 70 min). Basedon the definition of entropy, more uniform spatial distribution of lightintensity results in greater entropy value. Thus, the spectra of entropydecreased homogenously along the spectral wavelengths with the loss ofmoisture content and increasing drying time. At approximately 660 nm, adownward peak, which corresponds to the wavelength region of chlorophyll

absorption, was primarily observed ( Fig. 3 ).

3.3. Prediction models for color and moisture content of dried soybeans

Mean, entropy, and their combination were tested for predicting colorand moisture content using PLSR models. The color prediction results forthe three different feature sets are shown in Table 1 . Compared with theentropy method, mean reflectance achieved better results for calibration and

prediction models. The average correlation coefficient, R C, of 10 runs washigher by 1.7 %, whereas the average RMSEP values were reduced by 6.4%.For the prediction samples, the mean reflectance model improved theaverage correlation coefficient R P by 2.7% and the average RMSEP by8.8%. The paired t-test (p 6 0.05) showed that the PLSR model usingentropy had relatively lower prediction results for

Actual Color

Model LVs a R C b RMSEC c R CV b RMSECV c R P b RMSEP c,d RPD e

Mean Reflectance 11 0.909 0.87 0.881 0.99 0.862 1.04 A 2.0 Entropy 19 0.894 0.93 0.851 1.10 0.839 1.14 B 1.8 Mean Reflectance Combined Entropy 11 0.908 0.87 0.881 0.99 0.861 1.04 A 2.0

a Number of LVs.

b R C, R CV, and R P: correlation coefficient of calibration, cross validation, and prediction, respectively. c

RMSEC, RMSECV, and RMSEP: root-mean-square err or of calibration, cross validation, and prediction, d Different letters (A, B) for the paired RMSEPs in the column indicate significant differences (p 6 0.05) b e

Ratio of the standard error of performance to the standard deviation of the reference data.

Table 2 Average of 10 calibration and prediction results for moisture content by PLSR models f or soybeans.

espectively. etween prediction m

odels.

Model LVs a R C b RMSEC c (%) R CV b RMSECV c (%) R P b RMSEP c,d (%) RPD e

Mean Reflectance 15 0.984 3.5 0.974 4.6 0.971 4.7A 4.3 Entropy 17 0.954 5.9 0.928 7.5 0.901 9.2B 2.1 Mean Reflectance Combined Entropy 16 0.983 3.7 0.972 4.7 0.973 4.6A 4.4

a Number of LVs. b R C, R CV, and R P: correlation coefficient of calibration, cross validation, and prediction, respectively.

c RMSEC, RMSECV, and RMSEP: root-mean-square error of calibration, cross validation, and prediction, respectively. d Different letters (A, B) for the

paired RMSEPs in the column indicate significant differences (p 6 0.05) between prediction models. e Ratio of the standard error of performance to the standard deviation of the reference data.


6/7


Actual Moisture Content (%)

Fig. 5. Prediction of the color (a) and moisture content (b) of soybeans usingPLSR models coupled with mean reflectance. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web versionof this article.)

Wavelength (nm)

Wavelength (nm)

Fig. 6. Loadings on LV curves for color (a) and moisture content (b) using meanreflectance. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

color (R P = 0.839 and RMSEP = 1.14) than that using meanreflectance (R P = 0.862 and RMSEP = 1.04). Moreover, theintegration of mean reflectance and entropy data in a regressionmodel were as effective in color prediction as using the meanreflectance data alone. No significant statistical difference was found

between these methods (p-value


7/7


Latent Variable

Fig. 7. RMSECV versus LV curves for color (a) and moisture content (b). (Forinterpretation of the references to color in this figure legend, the reader is referred tothe web version of this article.)

over fitting was not observed from RMSECV versus latent variablecurves ( Fig. 7 ) for the prediction of color and moisture content.According to Shenk et al. (2001) , a robust model can be achieved if

the corrected standard error of prediction for bias (RMSEP corrected )does not exceed 1.30 times the RMSEC and when the bias value(BV) does not exceed 0.6 times the RMSEC. In this study, the upperlimit values were 1.25 (RMSEP corrected /RMSEC) and 0.56(BV/RMSEC), respectively, which indicate that the developedmodels are robust.

Although the models using mean reflectance as features achievedthe best results for color and moisture content, more work has to bedone to improve prediction accuracy, particularly for color. Futurestudies would address wavelength-selection approaches to identifythe optimal wavelengths among the 94 wavelengths so as to removethe redundant information. Moreover, future works would focus onthe evaluation of other image analysis methods for feature extraction(such as Fourier, moments, and fractal analysis). Furtherimprovements would render the technique useful for practical

applications.

5. Conclusion

Two statistical image features, mean reflectance and entropy,were extracted from hyperspectral images of dried vegetablesoybeans in the wavelength range of 400 1000 nm and then testedfor predicting their color and moisture content using PLSR models.PLSR models with mean reflectance yielded good prediction forcolor, with R P = 0.862 and RMSEP = 1.04. Compared with the

predictions of color, PLSR achieved better results for moisturecontent (R P = 0.971, RMSEP = 4.7%). The research results indicatethat hyperspectral reflectance images over the wavelength range of400 1000 nm can be used to evaluate the color and moisture content

of soybeans simultaneously during drying.Acknowledgments

The authors gratefully acknowledge the financial support fromChina 863 HI-TECH R&D Program (No. 2011AA100802),

National Natural Science Foundation of China (Grant nos.61271384 and 61275155), Natural Science Foundation of JiangsuProvince ( China, BK2011148), Postdoctoral Science Foundationof China (Grant nos. 2011M500851 and 2012T50463), the 111Project (B12018) and PAPD of Jiangsu Higher EducationInstitutions. References

Aguilera, J.M., 2003. Drying and dried products under the microscope. Food Sc i.Technol. Int. 9 (3), 137 143.

Ariana, D.P., Lu, R., 2008. Quality evaluation of pickling cucumbers usi ng h yperspectralreflectance and transmittance imaging Part II. Performance of a prototype. Sens.Instrum. Food Qual. Saf. 2 (3), 152 160.

Chan, T.F., Vese, L.A., 2001. Active contours without edges. IEEE Trans. Ima ge P rocess. 10 (2),266 277.

Del Fiore, A., Reverberi, M., Ricelli, A., Pinzari, F., Serranti, S., Fabbri, A.A., Bonifazi, G .,Fanelli, C., 2010. Early detection of toxigenic fungi on maize by hyperspectr al i maginganalysis. Int. J. Food Microbiol. 144 (1), 64 71.

Faustino, J.M.F., Barroca, M.J., Guin, R.P.F., 2007. Study of the drying kinetics of g reen bell pepper and chemical characterization. Food Bioprod. Process. 8 5, 1 63 170 (3 suppl. C).

Fernndez, L., Castillero, C., Aguilera, J.M., 2005. An application of image analysis ofdehydration of apple discs. J. Food Eng. 67 (1), 185 193.

Hou, J., Wang, C., Hong, X., Zhao, J., Xue, C., Guo, N., Gai, J., Xing, H., 2011. Associati onanalysis of vegetable soybean quality traits with SSR markers. Plant Breedi ng 1 30 (4),444 449.

Hu, Q., Zhang, M., Mujumdar, A.S., Du, W., Sun, J., 2006a. Effects of different dryi ngmethods on the quality changes of granular edamame. Drying Technol. 24 (8 ), 1025 1032 .

Hu, Q., Zhang, M., Mujumdar, A.S., Xiao, G., Sun, J., 2006b. Drying of edamame by h ot a ir andvacuum microwave combination. J. Food Eng. 77 (4), 977 982.

Hu, Q., Zhang, M., Mujumdar, A.S., Xiao, G., 2007. Performance evaluation of vacu ummicrowave drying of edamame in deep-bed drying. Drying Technol. 25 (4 ), 731 736.

Huang, M., Lu, R., 2010. Apple mealiness detection using hyperspectral scatteri ng t echnique.Postharvest Biol. Tecnol. 58 (3), 168 175.

Huang, M., Wan, X., Zhang, M., Zhu, Q., 2013. Detection of insect-damag ed vegetablesoybean using hyperspectral transmittance image. J. Food En g. 1 16 (1), 45 49.

Li, J., Rao, X., Ying, Y., 2012. Development of algorithms for detecting citrus cank er based onhyperspectral reflectance imaging. J. Sci. Food Agr. 92 (1), 125 134.

Liu, Y., Chen, Y., Wang, C., Chan, D., Kim, M.S., 2006. Development of hyperspectr alimaging technique for the detection of chilling injury in cucumbers; spectr al and imageanalysis. Appl. Eng. Agric. 22 (1), 101 111.

Lucas, A., Andueza, D., Rock, E., Martin, B., 2008. Prediction of dry matter, Fat, p H,vitamins, minerals, carotenoids, total antioxidant capacity, and color in fre sh a nd freeze-dried cheeses by visible-near-infrared reflectance spectroscopy. J. Agr. Food Chem. 56(16), 6801 6808.

Mendoza, F., Dejmek, P., Aguilera, J.M., 2006. Calibrated color measurements of a griculturalfoods using image analysis. Postharvest Biol. Tecnol. 41 (3), 28 5 295 .

Nicola, B.M., Beullens, K., Bolelyn, E., Peirs, A., Saeys, W., Theron, K.I. , Lammertyn, J .,2007. Nondestructive measurement of fruit and vegetable quality by means of N IRspectroscopy: a review. Postharvest Biol. Tecnol. 46 (2), 99 118.

Peng, Y., Zhang, J., Wang, W., Li, Y., Wu, J., Huang, H., Gao, X., Jiang, W., 201 1. P otential prediction of the microbial spoilage of beef using spatially resolv ed hyperspectralscattering profiles. J. Food Eng. 102 (2), 163 169.

Romano, G., Argyropoulos, D., Nagle, M., Khan, M.T., Mller, J., 2012. Combination ofdigital images and laser light to predict moisture content and color of be ll peppersimultaneously during drying. J. Food Eng. 109 (3), 438 448.

Shahin, M.A., Symons, S.J., 2011. Detection of fusarium damaged kernels in Cana da w esternred spring wheat using visible/infrared hyperspectral imaging a nd p rinciple componentanalysis. Comput. Electron. Agr. 75 (1), 107 112.

Shenk, J.S., Workman, J.J., Westerhaus, M.O., Bums, D.A., Ciurczak, E.W., 200 1.Application of NIR spectroscopy to agricultural products. Pract. Spectrosc. Se r. 2 7, 419 474.

Wang, Y., Zhang, M., Mujumdar, S.A., Mothibe Joseph, K., 2013. Microwave-assist ed p ulse-spouted bed freeze-drying of stem lettuce slices-effect on produ ct quality. FoodBioprocess Tech. 6 (12), 3530 3543.

Wu, D., Yang, H., Chen, X., He, Y., Li, X., 2008. Application of image texture for t he s ortingof tea categories using multi-spectral imaging technique and suppo rt vector machine. J.Food Eng. 88 (4), 474 483.

Wu, D., He, Y., Nie, P., Cao, F., Bao, Y., 2010. Hybrid variable selection in visible a nd n ear-infrared spectral analysis for non-invasive quality determination of gra pe juice. Anal.Chim. Acta 659 (1 2), 229 237.

Zhu, Q., Feng, Z., Huang, M., Zhu, X., 2013. Maize seed identification usi ng h yperspectralimaging and SVDD algorithm. Spectrosc. Spect. Anal. 33 (2 ), 517 521.
http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0030http://refhub.elsevier.com/S0260-8774(13)00612-2/h0030http://refhub.elsevier.com/S0260-8774(13)00612-2/h0030http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0125http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0120http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0115http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0110http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0105http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0100http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0095http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0090http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0085http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0080http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0075http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0070http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0065http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0060http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0055http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0050http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0045http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0040http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0035http://refhub.elsevier.com/S0260-8774(13)00612-2/h0030http://refhub.elsevier.com/S0260-8774(13)00612-2/h0030http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0025http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0020http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0015http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0010http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005http://refhub.elsevier.com/S0260-8774(13)00612-2/h0005

Documents

1 - Color y Humedad de Soya Deshidratada Con Imagenes Hiperespectrales