h yp ers p ectralima g er yclassificationusin gthecombinationofimp rovedla p lacianei g enma p sandimp rovedk－nearestnei g hborclassifier

hyp ers p ectral ima g er y（ hsi） classificationcanbeachievedthrou g hthecombinationschemeofnonlineardimensionalit yreductionusin gla p lacianei g enma p s（ le）andclassificationusin gthek-nearestnei g hbor（ knn） classifier.however，boththeleandtheknnclassifieromits p atialfea-turesofhsidataastheima g er y.thatseriousl yrestrictstheclassificationresultofhsidata.thisp a p erp resentstheada p tivewei g htedsummationkerneldistance（ awskd） toimp roveboththeleandtheknnclassifier，aimin gtop romotetheclassificationaccuraciesofhsidata.theawskdconsidersthes p ectralands p atialfeaturesofhsidata，andada p tivel yestimatethep ro p ers p atialnei g hborhoodsizefordescribin gthes p atialfeatureofeachp ixel.theindianandpaviaudatasetsareutilizedtoanal y zeandtestif ythep ro p osedcombinationschemeofimp rovedle （ ile）andimp rovedknn （ iknn）classifier.ex p erimentalresultsshowthatthep ro p osedcombinationschemeachievesshar p l yhi g herclassificationaccuraciesthanthere g ularschemeofleandknn.