基于高光谱成像的蓝莓内部品质检测 特征波长选择方法研究

在特征波长选取方面有一些创新，可以作为参考。在特征波长选取方面有一些创新，可以作为参考。（基于高光谱成像的蓝莓内部品质检测特征波长选择方法研究古文君1 ，田有文 1* ，张芳1 ，赖兴涛 1 ，何宽1 ，姚萍1 ，刘博林 2）586-482016620010~15mm0.8~2.3g。fone3:(InSpector V10E, Spectral InFinland)1392pix×1040pixCCDL CCD2(IGV-B141OM, IMPERX Incorporated, USA), 150W1. CCD Camera; 2.Spectrometer; 3.Shot; 4. Light source; 5. Samples(3900 Illuminatior, Illumination Tech6.Translationplatform7.Lightsourcecontroller;8.computernologies inc.,USA)、(IRCP0076-19. Translation platform controllerCOM,)、(120cm×50cmx(DELL VoStro 5560D-1528Figure 1 Schematic diagram of hyperspectral imagingcmsystem400~1000nm,4722.8nmRRGY-4(10mm)(DBR45(successive projections algorithm, SPA(stepwise multiple linear regression, SMLR)(SPA)(SMLR)SPASPASMLRSPA-SPA、SMLR_SMLR、SPA- SMLRSMLR-SPA21994-2018ChinaAcadcmicJournalElcctronicPublishingHousc.Allrightsrcscrved.http://www.cnki.nct5871.6BP(error back propagation)BP17(correlation coeffiient of calibration, Re)(root mean square error of calibration set, RMSEC)correlation coeffiient of pre-diction, Rp)(root mean square error of prediction set, RMSEP)ENVI 4.8(Research System Inc, ), MATLAB 2014a(The Math Works Inc)、TheUnscrambler9.7、 Excel2010(Ⅵ icrosoftdgle banddWcvef.BP models for soluble solidsThe selected characteristic wavelengthCurve of relative reflectanceExtract the region of interescontent and firmness prediction2figure 2 Flow chart of data processing280mm,68ms,28mm·s-。99%202.2600nm600nm2b2c)21994-2018ChinaAcadcmicJournalElcctronicPublishingHousc.Allrightsrcscrved.http://www.cnki.nct5884823(2f)BPSavitzky-Golasavitzky -golayTable 1 The effect of different spectra preprocessingCalibration setPredictioSpectrum typeRMSECRMSEPOriginal spcctrum0.933/0.9230.3510.4040.9200.9100.508/0.319MSCThe spectrum after MSC processing0.940/0.9450.56lO.3120.9190.9320.516/0.282SNThe spectrum after SNV processin0.93709340.60210.24309220.9010.6320.462Savitzky-golayThe spectrum after Savitzky-Golay processing 0.955/0.9550.3240.2410.951/0.9490.400/0.2782.5SPA-SPA SMLRSMLR SPA-SMLR SMLR-SPASPA-SPASPASavitzky-GolaySPATable 2 The results of multi-stage characteristic wavelength selection methodnmCharacteristie wavelength selection methodSPA-SPA452,455,470,482,490,785,893,912,921,942,950455,470,482,785,893.912SMLR-SMLR457,508,516,534,543,51,556,568,712,720.774,778508,534,543,712,720,774SPA-SMLR452,455,470,482,490,785,893,912,921,942,950452,470,482,490,893,912SMLR-SPA457,508,516,534,543,551,556,568,712,720,774,78534,7202.6Savilzky-gola(FS)392SPA-SPASMLR-SMLRSMLR-SMLRSMLR-SPABPBP0.001500021994-2018ChinaAcadcmicJournalElcctronicPublishingHousc.Allrightsrcscrved.http://www.cnki.nct589BPBPSPA-SPARp RMseP0.9520.391°Brix,RpRMSEP0.9530.234BrixTable 3 Detection results of soluble solid content and firmness of blueberry based on different multi-stagecharacteristic wavelength selection methodsCalibration setPrediction setCharacteristic selection method Wavelength numberRMSECRMSEP3929550.9550.324/0.2410.9510.9490.400/0.278SPA-SPA0.9590.9560.3180.1530.9520.9530.391/0.234SMLR-SMLR0.9560.9340.414/0.243912109020.559/0.349SPA SMLR0.828/0.8581.3670.58582208091.440/0.719SMLR- SPA20.958/0.9360.402/0.3359320.9280.435/0,4041387nm1229nm91.5%BPRRMSEP0.904215.163lBP3Rv0.84V0.94Rv0.83,SEV0.63。400-1000nmSavitzky-GolayBPSPA-SPASPA-SPA21994-2018ChinaAcadcmicJournalElcctronicPublishingHousc.Allrightsrcscrved.http://www.cnki.nct59048[1 KADER F,ROVEL. B Fractionation and identification of the phenolic compounds of highbush blueberries(Vaccinium corymbosumLUJ].Food Chemistry, 1996,55(1): 35-40「J,2012,33(1):340-342,2017,38(2):301-305.[4 MENDOZA F, LU R, ARIANA D,et al. Integrated spectral and image analysis of hyperspectral scattering data for prediction ofple [ruil firmness and soluble solids conlenl[J] Poslharvesl Biology and Technology, 2011, 62(2: 149-160[5 SUN M J, ZHANG D, LIU L,et al. How to predict the sugariness and hardness of melons a near-infrared [J]. Food Chemistry,2017,218(3:413-42116 SIEDLISKA A, BARANOWSKI P, MAZUREK W, ct al. Classification models of bruise and cultivar detection on the basis of hy-perspectral imaging data[J]. Computers and Electronics in Agriculture, 2014, 106: 66-74[7 LIU D, SUN D W, ZENG X N, el al. Recenl aDvances in wavelength seleclion lechniques for hyperspectral image processing inthe food industry[J]. Food Bioprocess Technol, 2014, 7: 307-323[8 ZHANG C, GUO C T, LIU F,et al. Hyperspectral imaging analysis for ripeness evaluation of strawberry with support vector ma-chine[j] Journal of Food Engincering, 2016, 179: 11-18[9J,2016,47(5:634-6402009,29(:1611-1615201536(12)171-17612]J,2012,32(11:3093309[13] LI B C, HOU B L, ZHANG D W,et al. Pears characteristics (soluble solids content and firmness prediction, varieties) testingInethods based on visible-near infrared hyperspecTral imaging[J]. OpLik, 2016, 127: 2624-2630[14] FAN S X, ZHANG B H,LI J B, et al. Prediction of soluble solids content of apple using the combination of spectra and textural features of hyperspectral reflectance imaging data[J. Postharvest Biology and Technology, 2016, 121: 51-61[15 RAJKUMAR P, WANG N,EIMASRY G, et al.Studies on banana fruit quality and maturity stages using hyperspectral imaging[ JIJournal of Food Engineering 2012, 108: 194-200,2015,36(16):10172015,35(8:2297-2302[18]WANG N,2007,23(2:151-155.「192008,39(5):91-9320」201536(10:70-74.[21] WU D, SUN D WAdvanced applications of hyperspectral imaging technology for food quality and safety analysis and assess-ment a review part T[J]. Innovative Food Science and Emerging Technologies, 2013, 19(4): 1-14J2014,35(8:57-61BP,2012.124」13,44(2):142-146.25],201523(6:1530-1537M011:41-48.[27,2013,24(10:1972-19762010,30(10):2729-2733?1994-2018ChinaAcadcmicJournaleLcctronicPublishingHousc.Allrightsreservedhttp://www.cnki.nct