搜索
30220114J．InfraredMillim．WavesVol．30No．2April20111001－9014201102－0124－072010-05-242010-10-09Receiveddate2010-05-24reviseddate2010-10-09407011208632006AA12010840920162008B331984-E-mailcici5201@163．com．*E-maillicj@nercita．org．cn．LAI12122*22221．3100292．100097LAI．．．．PCALAITP79AEstimatingleafareaindexfromremotesensingdatabasedondatasegmentationandprincipalcomponentanalysisDONGYing-Ying12WANGJi-Hua12LICun-Jun2*YANGGui-Jun2SONGXiao-Yu2GUXiao-He2HUANGWen-Jiang21．InstituteofAgriculturalRemoteSensing＆InformationSystemApplicationZhejiangUniversityHangzhou310029China2．NationalEngineeringResearchCenterforInformationTechnologyinAgricultureBeijing100097ChinaAbstractAccordingtotheunsatisfactoryandlowerefficiencyofclassicalstatisticalmodelsinleafareaindexLAIesti-mationanewinversionmethodcombinedwithphenology-baseddatasegmentationandprincipalcomponentanalysiswasproposedinthispaper．Inthemethodprincipalcomponentsofspectraldataanddifferentialordifferencespectraldatawerechosenasindependentvariablesandphenology-baseddatasegmentationwasintegratedintodataprocessinginordertoimproveestimationaccuracy．Inadditionmulti-scalewasinvolvedinmodeling．Winterwheatwasselectedasexperi-mentalobjectfornumericalsimulationandcomparativeanalysis．Resultsnotonlyshowedhighprecisioninwholeestimationandeffectivelyimproveddatasaturationbutalsomanifestedstabilityandrobustnessunderfullscan．KeywordsprincipalcomponentanalysisPCAphenologydatasegmentationmulti-scalemodelingleafareaindexLAIPACS41．20．-qLeafAreaIndexLAI、1．LAI2．、、、、1．、、．、、LAI13～5．2LAI．67LAI．8、、．．11．12009211、19、9428、81410．2009429517．ASDASDFieldSpecProFR350～2500nm350～1000nm1．4nm1000～2500nm2nm．1．3m25°20．50cm×4．1．21．2．1350～1354nm5nm．HJ-11．、．PrincipalComponentAnalysisPCA1．HJ-11HJ-1CCDFig．1SpectralresponsecurveofresourcessatelliteHJ-1CCDcameras．482820．2/31/3．1n=48、n=32、n=16±30%±20%±30%±20%．1LAI、、Table1StatisticalparametersofLAIforallsamplescali-brationsamplesandvalidationsamples±30%±20%±30%±20%7．497．407．417．427．117．317．311．211．281．291．261．531．351．304．013．994．014．004．044．004．033．633．623．643．623．693．633．653．943．913．923．933．973．953．941．691．681．701．711．671．711．712．862．842．902．912．822．932．910．240．250．240．240．210．240．24－0．90－0．91－1．01－1．01－1．05－0．96－0．961．2．2LAI1．21～7．499．10．52130DataSegmentation11．LAI＜33＜LAI＜5LAI＞510．35．．HJ-1LAI．、、LAILAI．HJ-112．LAI21．2HJ-1．3NDVI、SAVI、ARVI、EVI．1．2．3、2．101．1．2Fig．2PlanofnumericalexperimentsNDVI、SAVI、ARVI、EVI1．3．2Table2ThreekindsofassessmentindexesRMSERMSE=∑ni=1di/n－1槡ndi．R2R2=∑ni=1xi－x2yi－y()[]2∑ni=1xi－x2∑ni=1yi－y()[]2xi、x、yi、y、、、、n．Accuracyy=Accuracy·xxy．3Table3FourkindsofvegetationindexesNDVINDVI=ρNir－ρRedρNir+ρRedSAVISAVI=1+LρNir－ρRedρNir+ρRed+LL=01ARVIARVI=ρNir－ρRBρNir+ρRBρRB=ρRed－γρBlue－ρRedγ=1EVIEVI=2ρNir－ρRedρNir+c1ρRed+c2ρBlue+Lc1=6c2=7.5L=1ρNir、ρRed、ρBlur、760～900nm、630～690nm450～520nm．6212LAI．LAI、ND-VILAI、NDVI12LAI．LAINDVI．3HJ-1．22．1、LAI．4．99．5%、64．9%、40．4%．3Fig．3Saturationpoint4M1/2/3/4/5Spectral、Differential-1、Differ-ential-2、Spectral+Differential、Spectral+Differenti-al+Datasegmentation2．M2M1M34、Fig．4Contributionandcumulativecontributionoffirstsevenprincipalcomponentsofgroundhyperspectralfirst-orderandsec-ond-orderdifferentialSpectral．M4M1、M2、M3．M5RMSE0R2Accu-racy1．M5．2．2HJ-1HJ-12．1．．．5HJ-M1/2/3/4HJ、HJ+Datasegmentation、HJ+PCA、HJ+PCA+Datasegmentation2．HJ-M2HJ-M1HJ-M4HJ-M3．HJ-M3HJ-M1．HJ-M4HJ-M2．HJ-1LAI．721304Table4Resultsofnumericalexperimentsbasedongroundhyperspectral±30%±20%±30%±20%±30%±20%M10．991．001．000．690．710．700．950．970．98M20．870．890．890．750．750．751．021．051．07M30．890．900．900．740．760．751．191．241．23M40．900．910．930．730．740．731．041．061．06M50．690．690．720．840．840．831．011．061．085HJ-1Table5ResultsofnumericalexperimentsbasedonHJ-1multispectral±30%±20%±30%±20%±30%±20%HJ-M11．021．041．050．700．710．700．930．950．95HJ-M20．750．710．730．840．850．840．900．950．94HJ-M31．021．031．040．690．700．700．910．930．93HJ-M40．700．630．640．860．870．870．961．021．005Fig．5Comparisonanalysisofgroundhyperspectralnumericalexperiments2．356HJ-1“○”“△”“□”“●”“▲”“■”NDVI、SAVI、ARVI、EVINDVI、SAVI、ARVI、EVI．78“☆”“★”．56RMSER2Accuracy1．5R213．1～45．8%RMSE22．1～48．8%．6R22．46～50．5%RMSE6．53～48．1%．RMSER2Accuracy1．5R23．69～46．2%RMSE1．0～13．7%．6R213．2～55.7%RMSE16．1～17．6%．78．8212LAI、．．56±30%±20%．12．2%．3921308HJ-1Fig．8ResultsofHJ-1multispectralnumericalexperiments、．、．LAI．．．．REFERENCES1WANGJi-HuaZHAOChun-JiangHUANGWen-Jiangetal．BasisandApplicationofAgricultureQuantitativeRemoteSensingM．BeijingSciencePress．．2008．2VerstraeteMMPintyBMyneniRB．Potentialandlimi-tationsofinformationextractionontheterrestrialbiospherefromsatelliteremotesensingJ．RemoteSensingofEnvi-ronment1996582201—214．3RondeauxGStevenMBaretF．Optimizationofsoil-ad-justedvegetationindicesJ．RemoteSensingofEnviron-ment199655295—107．4YANGFeiZHANGBaiSONGKai-Shanetal．Compari-sonofmethodsforestimatingsoybeanleafareaindexJ．SpectroscopyandSpectralAnalysis．．200828122951—2955．5WANGXiu-ZhenHUANGJing-FengLIYun-Meietal．ThestudyonhyperspectralremotesensingestimationmodelsaboutLAIofriceJ．JournalofRemoteSensing．．20048181—88．6WENCheng-LinZHOUDong-Hua．TheroyandApplicationofMultiscaleM．BeijingTsinghuaUn