基于支持向量機(jī)的SAR圖像分割.pdf

1、西安科技大學(xué)碩士學(xué)位論文基于支持向量機(jī)的SAR圖像分割姓名：李瓊申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：付燕20080415Subject：SARImageSegmentationBasedonSupportVectorMachineSpecialty：ComputerApplicationTechnologyName：LiQiongInstructor：FUYanABSTRACT(Signature)(Signature)SA

2、R，asallalternativemeansoflong—distance—to—earthobservation，withhighresolution。all。weathertheadvantagesoftransmission，inremotesensing，militaryhydrologyminingandotherfields，havebeenrapidlydeveloped，andthereforetheanalysiso

3、ftheSARimageprocessingisbecomingincreasinglyimportant，inwhichSARimagesegmentationisanimportantaspectofthisThecurrentsegmentationmethodsinthetreatmentofllighdimensionalfeaturessmallsampleareasaredifficulttoobtaingoodsegme

4、ntationresults，mainlyforthepoorperformanceofthepromotionofcapacityorlearningslowlydifficultconvergenceInaddition，SARisacoherentimagingsystem，anditiseasilyaffectedbyspecklenoiseInthispapersupportvectormachineisusedinSARim

5、agesegmentationWedividetheoriginalimageintoasmallpieceofNNimage(aregional)，thencalculatethethreelocalstatisticaltexturefeaturesandtheelevengrayscaleimageblockCOoccurrencematrixfeaturesofeachsub—regionalimageAllthevalueso

6、btainedarescaledinto[1，1】，thenaresenttoSVMtobetrainedastheprescribedformatastraininginputsampletogetthefinaltraininggoodclassifierontheSARimagesegmentationAnimprovedoneagainstoneapproachiSusedtoextendtwo—classSVMtoamulti

7、classSVMthusmultitargetSARimagesegmentationisachievedInthesingle，multitargetSARimagesegmentationprocess，weusedifferentparametersanddifferentfeaturecombinationstofindthebestclassifierSOthatwecangetbettersegmentationresult

8、sandlowererrorrateAlsoacomparisonofdifferentsegmentationapproachisusedtoshowthedifferenteffectofaSARimageTheanalysisshowsthatthismethodisbetterthanmanyexistingmethodsbothintheaccuracyandtheantinoiseKeywords：SupportVector