polsarima g eclassificationusin gfeaturefusional g orithm basedonfeatureselectionandbila y ersvm

sin g let yp eoffeaturevectorcannotfull ydescribeob j ects,inordertofull yusetherichob j ectinformationofp olarimetricsarima g esandsolvethisp roblem,thisp a p erp utforwardanovel featurefusional g orithmbasedonfeatureselectionandbila y ersvmforp olarimetricsarima g eclassification-thatcanmakefulluseofthecomp letenessanddissimilarit ybetweenthefeaturestoformamoreeffec-tivefeaturevector.varioust yp esoffeaturevectorswereextractedfromanori g inal ima g eb ydifferentmethodsforfull ydescribin gthepolsardata.thefeaturevectorswerenormalizedtoensureeachfea-turevectorcanbeselectedunderthesamestandardsandhavethesameroleinclassification.as p atialpy ramidisintroducedtog etthefeaturevectorindifferentsizeors p atiallocation.a mrmrfeatureselectionmethodwasusedtoobtaintheo p timalfeaturesubsetforg ivencate g oriestoavoidredundan-c yandoverfittin gp henomenoncausedb ythesimp lecombinationofvariousfeaturevectors.finall y，themultila y erconce p twasintroducedandabila y ersvm modelwasconstructedtoo p timizeandre-p rocessthep robabilitiesofthetar g etcate g or yobtainedb ythefirstsvm.ex p erimentalresultsonthetwop olarimetricsarima g esachievedb ythejetpro p ulsionlaborator yshowthesu p eriorit yofthep ro p oseda pp roach.