AimingtosolvethemisclassificationproblemsofunsupervisedpolarimetricWishartclassificationalgorithmbasedonFreemandecomposition,anunsupervisedPolarimetricSyntheticApertureRadar(SAR)Interferomery(PolInSAR)classificationalgorithmbasedonoptimalcoherencesetparametersisstudiedandproposed.ThisalgorithmusestheresultofFreemandecompositiontopidetheimageintothreebasiccategoriesincludingsurfacescattering,volumescattering,anddouble-bounce.Then,thePolInSARoptimalcoherencesetparametersareusedtofinelypideeachofthethreebasiccategoriesinto9categories,andthewholeimageispidedinto27categories.BecauseboththeFreemandecompositionresultandoptimalcoherencesetparametersindicatespecificscatteringcharacteristics,thewholeimageismergedinto16categoriesbasedonphysicalmeaning.Atlast,theWishartclusterisemployedtoobtainthefinalclassificationresult.Topreservethepurityofscatteringcharacteristics,pixelswithsimilarscatteringcharacteristicsarerestrictedtobeclassifiedwithotherpixels.Thefinalclassificationresultseffectivelyresolvethemisclassificationproblem,notonlythebuildingscanbeeffectivelydistinguishedfromvegetationinurbanareas,butalsotheroadiswelldistinguishedfromgrass.Inthispaper,theE-SARPolInSARdataofGermanAerospaceCenter(DLR),areusedtoverifytheeffectivenessofthealgorithm.
