极化干涉SAR森林高度反演方法研究

1007-4619(2010)04-806-16JournalofRemoteSensing遥感学报Received:2009-09-11;Accepted:2010-02-01Foundation:NationalNatureSciencesFoundation(No.2007CB714404).Firstauthorbiography:LUOHuanmin(1972),male.HereceivedmasterdegreefromtheSchoolofAutomationEngineering,UniversityofElec-tronicScienceandTechnologyofChina(UESTC).HeisnowstudyingforPh.D.DegreeinInstituteoftheGeo-spatialInformationScienceandTechnology,UESTC.HisresearchinterestsaremainlythepolarimetricandinterferometricSARremotesensingforforestapplications.E-mail:[e-mail]nmethodsusingpolarimetricSARinterferometryLUOHuanmin1,CHENErxue2,CHENGJian3,LIXiaowen1,41.InstituteofGeo-spatialInformationScienceandTechnology,UniversityofElectronicScienceandTechnologyofChina,SichuanChengdu610054,China;2.InstituteofForestResourcesInformationTechniqueofChineseAcademyofForestry,Beijing100091,China;3.SchoolofElectronicEngineering,UniversityofElectronicScienceandTechnologyofChina,SichuanChengdu610054,China;4.ResearchCenterforRemoteSensingandGIS,SchoolofGeography,BeijingNormalUniversity,Beijing100875,ChinaAbstract:ForestheightextractionwithpolarimetricSARinterferometry(POLInSAR)isahotresearchfieldofimagingSARremotesensing.SeveralavailableforestheightinversionmethodsusingPOLInSARdatawerevalidatedandcomparedwithre-peatpassE-SARdatasetsandthecorrespondinggroundmeasuredforeststandheightthroughtheanalysisoftheRandomVol-umeoverGround(RVoG)scatteringmodel.Afteranalyzingtheexperimentresultsintheviewofphysicalmechanisms,wede-velopedanintegratedinversionmethodcombininginterferometriccoherenceoptimizationandcompensationofnon-volumetricscatteringdecorrelation.Validationresultshowsthatthegeneralperformanceofthedevelopedforestheightinversionmethodissuperiortotheothers.Keywords:POLInSAR,polarimetricinterferometriccoherenceoptimization,RVoG,forestheight,non-volumetricscat-teringdecorrelationCLCnumber:TP722.6/S7Documentcode:A1INTRODUCTIONTheforestheightisanimportantforestresourceinformationparameterandusuallyusedinbiomassestimationmodel.ForestheightextractionwithpolarimetricSARinterferometry(POLInSAR)isahotresearchfieldofimagingSARremotesensing.SARinterferometryisawell-establishedSARtech-niquetoestimatetheverticallocationoftheeffectivescatteringcenterineachresolutioncellthroughthephasedifferenceinimagesacquiredfromspatiallyseparatedantennas.Scatteringpolarimetryissensitivetotheshape,orientationanddielectricpropertiesofscatterers.POLInSARbasedonthecoherentcombinationofradarinterferometryandpolarimetryallowsustoovercometheseverelimitationsofbothtechniqueswhentakenaloneandisbecominganimportanttechniqueforforestparametersextraction(Wuetal.,2007).Cloudeetal.(2003)haveproposedthethree-stagesinversionalgorithmwhichisbasedonacoherentmixturemodelofarandomvolumeovergrounddescribingtherelationofforestheightandpolarimetricinterferometriccoherence.Atwo–componentpolarimetricin-terferometricmodelispresentedforimprovementofvegetationparameterretrievalusingtheNelder–Meadsimplexoptimalmethod.Itcombinesscatteringmodelbasedpolarimetricdecom-positiontechniqueandRVoGbasedPOLInSARforestparame-tersinversionmodel(Neummanetal.,2009).Hajnseketal.(2009)discussedtheeffectofsomefactors,suchastemporaldecorrelationandtopographyetc.,toforestheightinversionaccuracyusingP-,L-andX-bandairbornePOLInSARdataofIndonesia’stropicalforests.TheerrorsourcesofPOLInSARforestheightinversionhasbeenqualitativelyanalyzedusingSIR-C/XSARL-bandrepeatpassPOLInSARdataandthecorrespondingopticalimageac-quiredinHetianofXinjiang(Chenetal.,2007).SixforesttreeheightinversionmethodsofPOLInSARwerevalidatedusingrepeatpassE-SARdatasetsandthecorrespondinggroundmeasuredforeststandheight(Chenetal.,2008).Utilizingthedifferencesamongthepowersofbackwardscatteringsignalandscatteringcenterswithdifferentscatteringmechanisminthesameresolvingunit,somescholarsproposedTLS-ESPRITalgorithmtoextractdominantscatteringcenterphasesforforesttreeheightinversion.Themethodcanimprovecomputationalefficiency,butthecapabilityforimprovinginversionaccuracyislimited(Yangetal.,2007;Zhouetal.,2008).InordertoCitationformat:LuoHM,ChenEX,ChengJandLiXW.2010.ForestheightestimationmethodsusingpolarimetricSARinterferometry.JournalofRemoteSensing.14(4):806821LUOHuanminetal.:ForestheightestimationmethodsusingpolarimetricSARinterferometry807improvetheaccuracyofforesttreeheightestimation,Lietal.(2005)andChenetal.(2008)carriedoutrelevantresearchworktryingtoincreasetheutilizationofbaselineorfrequencyin-formationinobservationspace.Afteranalyzingtheexperimentresultsofseveralavailableforestheightinversionmethodsintheviewofphysicalmecha-nisms,aRVoGmodelbasedinversionmodelcombiningcoher-entamplitudewithphaseinformationisdevelopedinthispaper,wherethePhaseDiversity(PD)interferometriccoherenceop-timizationmethod(Tabbetal.,2002)isusedtoobtainthesur-faceandvolumescatteringphase.Inordertoinvestigatebetterinversionmethods,theperformanceoftheproposedmethodiscomparedwiththeotheravailablemethodsusingrepeatpassE-SARLbandPOLInSARdataandthecorrespondinggroundmeasuredforeststandheight.2TESTSITESANDDA