基于微多普勒效应的地面运动目标噪声稳健分类的研究

《基于微多普勒效应的地面运动目标噪声稳健分类的研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、II基于微多普勒效应的地面运动目标噪声稳健分类研究万方数据AbstractIIIAbstractMicro-Dopplereffectreflectsthegeometricalstructuresandthemovingcharacteristicsofthetargets,therefore,itisanewapproachforradarautomatictargetrecognition(ATR),whichofferstheeffectivetechniquesupportforone’sactivepositioninthebattle.However,du

2、etosomeseveremeasurementconditions,i.e.,thefardistancebetweenthetargetandradar,thehighSignal-Noise-Ratio(SNR)conditioncannotbeguaranteedforthemeasureddata.ThusnoiserobustnessbecomesanimportantjobinATRfield.Onthebasisofthepreviouswork,thisthesisstudiesthenoise-robustclassificationofthegro

3、undmovingtargetsviamodificationofextractedfeaturesandnoise-reduction.Themainworkisasfollows:(1)Thethesisintroducestheconceptsofmicro-Doppler,andestablishesthemicro-Dopplermodelformovingwheeledandtrackedvehiclesaswellasanalysesthedifferencesofmicro-Dopplereffectbetweenamovingvehicleandawa

4、lkinghumanbasedonmeasureddata.Inaddition,thethesisstudiestheexistingcluttersuppressionmethodsforthegroundmovingtargets.(2)Aftermakingananalysisonthedifferenceinthedistributionofharmoniccomponentsbetweenwheeledandtrackedvehicles,weintroducetwoclassificationmethodsbasedontheenergydistribut

5、ion.Sincethetargetandnoisesubspacesareunrelated,anoisereductionschemeisproposed.Theeffectivenessoftheproposedmethodisverifiedbyclassificationexperimentsbasedonmeasureddata.(3)A3-dimensionalfeaturevectordescribingthetime-varyingcharacterofinstantaneousfrequencyisextractedfromthetime-frequ

6、encyspectrogramofmicro-Dopplersignaturestodistinguishthemovingvehicleandwalkinghuman.ConsideringtheworseclassificationperformanceunderthelowSNRcondition,anewnoisereductionmethodisproposedbasedonsignalreconstructionwiththeBayesianInferenceCriterion(BIC)basedComplexProbabilisticPrincipalCo

7、mponentAnalysis(CPPCA)model,whichisextendedfromreal-valuedProbabilisticPrincipalComponentAnalysis(PPCA),andtheprincipalcomponentsareselectedthroughBIC.ComparedwiththeexistingCLEANandPrincipalComponentAnalysis(PCA)methodsfordenoising,theBICbasedCPPCAmethodpresentsitsadvant