hyperspectral imaging for food quality analysis and control || index
TRANSCRIPT
Index
ACA, see Axial chromaticaberration
Acousto-optic tunable filter(AOTF)
light sources, 137–138wavelength dispersion, 144–146,
457Adaptive thresholding, image
segmentation,110–111
ANN, see Artificial neural networkAOTF, see Acousto-optic tunable
filterApple
bruise damage
causes, 295–296hyperspectral imagingalgorithms for bruisedetection, 303–305,310–311, 313–315
cameras, 301–302illumination unit, 302imaging spectrograph,
299–301preprocessing of images,
302–303, 307sample preparation and
system setup, 305–306spectral characteristics of
normal and bruisedsurfaces, 311–313
wavelength selection, 303,307–310
traditional detectionmethods, 297–299
grading, 296market, 295
Area scanning, see Staring imageArtificial neural network (ANN)
apple bruise detection, 305
back propagation neural network,462–463
hyperspectral imageclassification, 91–92
meat quality assessment, 203ASCC, see Average squared
canonical correlationAutomation, importance in quality
assessment, 4–5Average squared canonical
correlation (ASCC),461
Axial chromatic aberration (ACA),465
Back propagation neural network(BPNN), 462–463
Band Interleaved by Line (BIL),132
Band Interleaved by Pixel (BIP),132
Band number, 19Band Sequential (BSQ), 132Bandpass filter, 143Bandwidth, 19, 144Beef, see Meat quality assessmentBIL, see Band Interleaved by LineBIP, see Band Interleaved by PixelBPNN, see Back propagation
neural networkBSQ, see Band Sequential
CA, see Correlation analysisCalibration, hyperspectral imaging
instrumentation
flat-field correction, 164–165radiometric calibration,166spatial calibration, 159–161spectral calibration, 161–164
overview, 32–36preprocessing
overview, 37, 45–46radiometric calibration
normalization, 65overview, 55–56percentage reflectance,
56–63relative reflectance
calibration, 63–64transmittance image
calibration, 64wavelength calibration
imaging system, 48–50purpose, 46technique, 50–55
reflectance calibration, 35Candling, nematode detection in
fish fillets, 215CART, see Classification and
regression treeCCD, see Charge-coupled deviceCharge-coupled device (CCD)
architectures, 153–154low light cameras, 156–158on-line poultry inspection
systems, 246–247,253–257
overview, 28, 31performance parameters,
154–156sensor materials, 153
Chemometrics, data analysis, 38Chicken
quality assessment withhyperspectral imaging
automated systemdevelopment
charge-coupled devicedetector, 246–247
471
Index472
Chicken (continued )
laboratory-basedphotodiode arraydetection systems,245–246
pilot-scale system, 246spectral classification, 248
contamination detection,220–227
line-scan imaging for on-linepoultry inspection
commercial applications,266–267
hyperspectral imaginganalysis, 257–261
in-plant evaluation,262–266
multispectral inspection,261–262
spectral line-scan imagingsystem, 255–257
on-line inspection, 229–230overview, 220target-triggered imaging
system developmentdual-camera and color
imaging, 249–250multispectral imaging
systems, 252–255two-dimensional spectral
correlation and colormixing, 250–252
tumor and disease detection,227–229
United States poultry inspectionprogram, 243–245
Chromatic aberration, 465Circular variable filter (CVF), 152Citrus fruit
defects, 321–322hyperspectral imaging
automated rotten fruitdetection, 339–344
hardware, 330illumination system, 328–330integration time correction at
each wavelength,331–333
overview, 326–328spatial correction of intensity
at light source, 333–334spherical shape corrections,
334–339
market, 321multispectral identification of
blemishes, 323–325Classification and regression tree
(CART), 255,342–343
CMOS, see Complementary metaloxide semiconductor
Color, meat quality assessment,179, 205
Complementary metal oxidesemiconductor(CMOS), cameras, 31,158–159
Computer vision systemadvantages and limitations, 5–6meat quality assessment,
183–184wheat classification, 455
Convolution, see Imageenhancement
Correlation analysis (CA), citrusfruit analysis, 341
Cucumberclassification, 431–432damage, 432–433hyperspectral imaging of pickling
cucumbers
bruise detection, 433–438internal defect detection, 438prospects, 445production, 431–432CVF, see Circular variable filter
DA, see Discriminant analysisDARF, see Directional average
ridge followerDark current, subtraction, 66DASH, see Digital array scanned
interferometerDatacube, 20Derivative filtering, image
enhancement,103–104
Digital array scanned interferometer(DASH), 152
Directional average ridge follower(DARF), fish qualityassessment, 215
Discriminant analysis (DA), 38Discriminant partial least squares
(DPLS), 219
DPLS, see Discriminant partialleast squares
ECHO, see Extraction andclassification ofhomogeneous objects
Edge-based segmentationedge detection, 112–113edge linking and boundary
finding, 114Electromagnetic spectrum, 14–15Electron-multiplying charge-
coupled device(EMCCD), 156–157,255–257
EMCCD, see Electron-multiplyingcharge-coupled device
Enhancement, see Imageenhancement
ENVI, see Environment forVisualizing Images
Environment for VisualizingImages (ENVI), imageprocessing, 119, 121
Essential wavelength, dataanalysis, 38
Extraction and classification ofhomogeneous objects(ECHO), 396–397
Factorial analysis, 465FDA, see Fisher’s discriminant
analysisFecal contamination, detection on
chicken, 220–227Filter wheel, 143–144Fish
quality assessment withhyperspectral imaging
freshnessidentification with
subjective region ofinterest, 277–282
morphometricsuperimposition fortopographical freshnesscomparison, 282–287
overview, 205–206, 273–277qualitative measurements,
210–220quantitative measurements,
206–210
Index 473
traditional quality assessment,273–274
Fish, see Meat quality assessmentFisher’s discriminant analysis
(FDA), 84–86Flat-field correction, 164–165FLIM, see Fluorescence lifetime
imaging microscopyFluorescence lifetime imaging
microscopy (FLIM), 10Focal plane scanning, see Staring
imageFourier transform
image enhancement
high-pass filtering, 106low-pass filtering, 105–106imaging spectrometers, 148–150Full width at half maximum
(FWHM), bandwidth,19, 144, 200, 252
FWHM, see Full width at halfmaximum
GA, see Genetic algorithmGabor filter, texture
characterization,117–118, 120
Gaussian kernel, 94Gaussian Mixture Model (GMM),
hyperspectral imageclassification, 80,89–91
Gel electrophoresis, wheatclassification, 453
Genetic algorithm (GA), citrus fruitanalysis, 342
GLCM, see Graylevelco-occurrence matrix
Global thresholding, imagesegmentation, 110
GMM, see Gaussian MixtureModel
Graylevel co-occurrence matrix(GLCM)
meat quality assessment, 195,198
texture characterization, 116–117
HACCP, see Hazard analysiscritical control point
Halogen lamp, light sources,133–134
Hazard analysis critical controlpoint (HACCP), 6, 24
Hemoglobin, fish qualityassessment, 214
High-performance liquidchromatography(HPLC)
compound distributionmeasurement inripening tomatoes,379–380, 383
wheat classification, 453–454Histogram equalization, image
enhancement,100–102
HPLC, see High-performanceliquid chromatography
HSI, see Hyperspectral imagingHypercube, 20–23Hyperspec, image processing,
122–123Hyperspectral imaging (HSI)
acquisition modes, 24–28,131–132
advantages, 3, 7–8calibration, see Calibration,
hyperspectral imagingcomparison with imaging and
spectroscopy,6–7, 130
components of system, 29–32disadvantages, 9–11fruit and vegetable analysis, see
Apple; Citrus fruit;Cucumber; Melon sugardistribution; Mushroom;Tomato
image classification, see Imageclassification
image data, 20–24image processing, see Image
enhancement; Imagesegmentation; Objectmeasurement
instrumentation
detectors, 28, 152–159light sources, 133–139wavelength dispersion devices,139–152meat, see Meat quality
assessmentsoftware, 118–123spectral data analysis, 36–39
synonyms, 6wheat kernels, see Wheat
ICCD, see Intensified charge-coupled device
ICM, see Iterated conditional modeIDA, see Independent component
analysisImage classification
artificial neural networks, 91–92Gaussian Mixture Model, 80,
89–91optimal feature and band
extraction
combination principalcomponent analysis andFisher’s discriminantanalysis, 85–86
feature search strategy, 82–83feature selection metric,
81–82Fisher’s discriminant analysis,
84–85independent component
analysis, 86–88principal component analysis,
83–84overview, 79–80support vector machine, 92–94
Image enhancementhistogram equalization, 100–102overview, 100spatial filtering
arithmetic operations, 109convolution, 102derivative filtering, 103–104Fourier transform
high-pass filtering, 106low-pass filtering, 105–106
median filtering, 103pseudo-coloring, 107–109smoothing linear filtering,
102–103wavelet thresholding,
105–106
Image segmentationedge-based segmentation
edge detection, 112–113edge linking and boundaryfinding, 114morphological processing,
111–112
Index474
Image segmentation (continued )overview, 109spectral image segmentation,
114–115thresholding
adaptive thresholding,110–111
global thresholding, 110
Imaging spectrograph, 32, 139–142Imaging spectroscopy, seeHyperspectral imagingImSpector V10E imaging
spectrograph, 141,160, 162
Independent component analysis(IDA), 86–88,383–385, 465
Intensified charge-coupled device(ICCD), 156–158
Iterated conditional mode (ICM),396
Kernel visual distinguishability(KVD), 451
KVD, see Kernel visualdistinguishability
Laser, light sources, 136–137Lateral chromatic aberration
(LCA), 465LCA, see Lateral chromatic
aberrationLCTF, see Liquid crystal tunable
filterLDA, see Linear discriminant
analysisLED, see Light emitting diodeLight
characteristics, 13–14electromagnetic spectrum, 14–15interaction with samples, 16–18
Light emitting diode (LED), lightsources, 134–136, 458
Light sourceshalogen lamps, 133–134lasers, 136–137light emitting diodes, 134–136,
458tunable sources, 137–139
Line-scan imaging, see PushbroomLinear discriminant analysis (LDA)
citrus fruit analysis, 342–344
tomato maturity, 373–374,377–378
Linear variable filter (LVF), 152Liquid crystal tunable filter (LCTF),
146–148Luminosity value, see L-valueL-value, mushroom grading,
403–404, 425LVF, see Linear variable filterLycopene, see Tomato
Machine vision, see Computervision system
MATLAB, image processing,121–122, 464
Meat quality assessmentcolor, 179computer vision, 183–184destructive measurements,
179–182hyperspectral imaging
applicationsbeef, 194–202chicken, see Chickenfish, see Fishpork, 202–205
chemical imaging, 187–189data exploitation, 189–192overview, 185–186techniques, 192–193
objective technique assessment,182–183
overview, 175–177purpose, 178–179spectroscopy, 184–185standards, 178
Median filtering, imageenhancement, 103
Melon sugar distributionimaging spectroscopy
half-cut melon, 353instrumentation, 352–353intensity conversion to sugar
content, 354–355noise correction, 353–354partial image for sugar content
calibration, 353sugar absorption band
calibration, 362–363image acquisition, 362instrumentation, 360–361visualization, 364–365
sugar distributionvisualization, 355–356
melon features for study, 350near infrared spectroscopy
sample preparation, 350, 357sugar absorption band
calibration, 359–360data acquisition and sugar
content, 357–358second-derivative
spectrum, 358–359wavelength selection,
350–351overview, 349
MEMS, see Micromechanicalsystems
MI, see Mutual informationMicromechanical systems
(MEMS), 152Minimum noise fraction (MNF),
transformation,70, 304
Moisture content, mushrooms,423–424
Morphological processing, imagesegmentation,111–112
Multiplicative scatter correction(MSC), 408–410
Multispectral imagingcitrus peel blemishes, 323–325overview, 23poultry, 252–255, 261–262
Multivariate image analysis (MVI),464–465
Mushroombrowning and bruising,
403–404color vision, 404–405hyperspectral imaging
curvature and spectralvariation, 407–410
equipment, 405–407image classification
model building, 410–413regression models,
420–427supervised classification
for freezing injurydetection, 416–420
unsupervised classificationfor surface damagedetection, 413–416
Index 475
overview, 405prospects, 427–428sliced mushroom quality
attributes, 420–423whole mushroom quality
attributescolor prediction, 425–427moisture content, 423–424
L-value in grading, 403–404,425
market for Ireland whitemushrooms, 403
spectroscopy, 403–404Mutual information (MI), citrus
fruitanalysis, 341–342MVI, see Multivariate image
analysis
Near infrared spectroscopy (NIRS)cucumber bruise detection,
434–437meat quality assessment, 183,
185, 195–196, 229multispectral identification of
citrus peel blemishes,323–325
principles, 6, 12–13wheat classification, 454
Nematodes, detection in fish fillets,215–220
NIRS, see Near infraredspectroscopy
Noise reduction, see Preprocessing
Object measurementintensity-based measures,
115–116relative reflectance equation,
115texture
Gabor filter, 117–118, 120graylevel co-occurrence
matrix, 116–117
Offner imaging spectrograph,141–142OPD, see Optical path distanceOptical path distance (OPD),
148–149
Partial least squares (PLS), 10,191, 207, 308–309,379–380, 413, 424
Partial least squares-discriminantanalysis (PLS-DA)
cucumber evaluation, 440fish freshness analysis, 279,
281–282, 285, 287mushroom evaluation, 411,
416–420PCA, see Principal component
analysisPCR, see Polymerase chain
reaction; Principalcomponent regression
pH, meat quality assessment, 205Phenol test, wheat classification,
453Pickle, see CucumberPlanck’s relation, 14PLS, see Partial least squaresPLS-DA, see PLS-DAPoint-scan imaging, see
WhiskbroomPolymerase chain reaction (PCR),
wheat classification,454
Polynomial kernel, 94Pork, see Meat quality assessmentPoultry, see ChickenPoultry Product Inspection Act
(PPIA), 243PPIA, see Poultry Product
Inspection ActPreprocessing
apple bruise detection, 302–303,307
calibration
radiometric calibrationnormalization, 65overview, 55–56percentage reflectance,
56–63relative reflectance
calibration, 63–64transmittance image
calibration, 64wavelength calibration
imaging system, 48–50purpose, 46technique, 50–55
noise reduction and removaldark current subtraction, 66minimum noise fraction
transformation, 70noisy band removal, 69–70
Savitzky–Golay filtering,67–69
spectral low pass filtering, 67overview, 37, 45–46
Principal component analysis(PCA)
cucumber quality evaluation forpickling, 435, 437, 459,465
image classification, 38, 79,83–86
meat quality evaluation
beef, 197–198chicken, 228overview, 191pork, 203–204mushroom quality evaluation,411, 414–416, 418–419
tomato ripening analysis,383–385
Principal component regression(PCR), 10, 413,421–422
Prism-grating-prism imagingspectrograph,139–141
Pseudo-coloring, imageenhancement,107–109
Pushbroom, 25, 27–28, 131–132,456
Quartz–tungsten–halogen lamp,133
Radiometric calibration, seeCalibration,hyperspectral imaging
Raster-scanning imaging, seeWhiskbroom
RDLE, see Refreshed delayed lightemission
Reflectance calibration, 35Refreshed delayed light emission
(RDLE), 433Relative prediction deviation
(RPD), 424, 426Ripening, see Melon sugar
distribution; TomatoRMSECV, see Root mean square
error of cross-validation
Index476
RMSEP, see Root mean square errorof prediction
Root mean square error of cross-validation (RMSECV),421, 427
Root mean square error ofprediction (RMSEP),380, 421, 427
RPD, see Relative predictiondeviation
Savitzky–Golay filtering, noise,67–69
SBFS, see Sequential backwardfloating selection
SBS, see Sequential backwardselection
SEE, see Standard error of estimateSegmentation, see Image
segmentationSequential backward floating
selection (SBFS),feature searchstrategy, 83
Sequential backward selection(SBS), feature searchstrategy, 82–83
Sequential forward floating selection(SFFS), feature searchstrategy, 83
Sequential forward selection (SFS),feature searchstrategy, 82–83
SFFS, see Sequential forwardfloating selection
SFS, see Sequential forwardselection
SG-FCM, see Spatially guidedfuzzy C-means
Shortwave near infrared spectralcamera, fish qualityassessment, 210
Sigmoid kernel, 94Signal-to-noise ratio (SNR), 19–20Single shot hyperspectral imagers,
150–152Slice shear force (SSF), meat quality
assessment, 181–182,194
Smoothing linear filtering, imageenhancement,102–103
SNR, see Signal-to-noise ratioSNV, see Standard normal variateSpatial filtering, see Image
enhancementSpatially guided fuzzy C-means
(SG-FCM), 396Spatial resolution, 19SpectraCube, image processing,
122–123Spectral image segmentation,
114–115Spectral low pass filtering, noise, 67Spectral range, 18Spectral resolution, 18–19Spectral signature, 20Spectrograph, see Imaging
spectrographSpectroscopy
hyperspectral imagingcomparison, 6–7, 130
principles, 11–13SSF, see Slice shear forceStandard error of estimate (SEE), 55Standard normal variate (SNV),
408–409, 424Staring image, 24–26, 131–132, 456Stepwise multivariate regression
(SW), citrus fruitanalysis, 342
Sugar distribution, see Melon sugardistribution
Support vector machine (SVM),hyperspectral imageclassification, 80,92–94, 460
SVM, see Support vector machineSW, see Stepwise multivariate
regression
Tenderness, meat qualityassessment, 179–182
Thresholding, see Imagesegmentation
Tomatocolor imaging of maturity,
371–372compound distribution
measurement inripening tomatoes,379–381
health benefits, 369hyperspectral imaging of maturity
combining spectral and spatialdata analysis
integrated spectral andspatial classifiers,396–398
overview, 390parallel spectral and spatial
classifiers, 391–395sequential spectral and
spatial classifiers, 391comparison with color
imaging, 375–376image acquisition, 373linear discriminant analysis,
373–374, 377–378normalization of images,
376–377preprocessing, 373prospects, 398–399spectral data classification,
377–379spectral data reduction,
387–390market, 369on-line unsupervised
measurement ofmaturity, 382–387
optical properties, 370–371ripening process, 370
Tumors, detection on chicken,227–228
Tunable filter scanning, see Staringimage
Ultraspectral imaging, 23–24
Variable importance in projection(VIP), 309
VHIS, see Volume holographicimaging spectrometer
VIP, see Variable importance inprojection
Volume holographic imagingspectrometer (VHIS),152
Warner–Bratzler shear force(WBSF), meat qualityassessment, 181–182,184, 194, 202
Water holding capacity (WHC),meat, 179, 205
Index 477
Wavelength calibration, seeCalibration,hyperspectral imaging
Wavelength difference, 437–438Wavelength ratio, 437–438Wavelength scanning, see Staring
imageWavelet thresholding, image
enhancement,105–106
WBSF, see Warner–Bratzler shearforce
WHC, see Water holding capacityWheat
applications, 449
classificationcomputer vision system, 455gel electrophoresis, 453high-performance liquid
chromatography,453–454
near-infrared spectroscopy,454
overview, 449–452phenol test, 453polymerase chain reaction,
454visual identification, 452
hyperspectral imaging forclassification
Canadian wheat classificationand accuracy, 462–463
challenges, 464–465detectors, 456–457hardware and software
integration, 458illumination sources, 458image classification, 459–461prospects, 465–466system types, 455–456vitreous versus non-vitreous
kernels, 461wavelength filtering devices,
457–458
Whiskbroom, 24–27, 131–132