预览加载中，请您耐心等待几秒...

基于局部特征的在轨卫星遥感图像舰船检测.docx

预览
1/2
2/2

在线预览结束，喜欢就下载吧，查找使用更方便

下载提示 文本预览

如果您无法下载资料，请参考说明：

1、部分资料下载需要金币，请确保您的账户上有足够的金币

2、已购买过的文档，再次下载不重复扣费

3、资料包下载后请先用软件解压，在使用对应软件打开

基于局部特征的在轨卫星遥感图像舰船检测 Title:ShipDetectioninOn-OrbitSatelliteRemoteSensingImagesBasedonLocalFeatures Abstract: Satellite-basedshipdetectionplaysacrucialroleinnumerousmaritimeapplications,includingmarinemonitoring,surveillance,andmaritimesecurity.Inthispaper,weproposeashipdetectionmethodbasedonlocalfeaturesinon-orbitsatelliteremotesensingimages.Themethodincorporatesamulti-stageapproachinvolvingfeatureextraction,featurematching,andshipclassification.Experimentalresultsdemonstratetheeffectivenessandrobustnessoftheproposedmethodindetectingshipsaccurately. 1.Introduction Shipdetectioninsatelliteremotesensingimageshasgainedmuchattentionduetoitspotentialapplicationsinmaritimedomainawareness.Traditionalshipdetectionmethodsrelyonglobalfeaturesandoftensufferfrominterferencecausedbycomplexbackgroundsandvaryingseastates.Thispaperproposesanovelshipdetectionmethodthatutilizeslocalfeaturestoaddressthesechallengesandimprovedetectionaccuracy. 2.FeatureExtraction Tocapturelocalfeatures,weemploytheScale-InvariantFeatureTransform(SIFT)algorithm.SIFTextractsdistinctivekeypointsandtheircorrespondingdescriptorsfromthesatelliteimages.TheSIFTalgorithmisrobusttochangesinscale,rotation,andillumination,makingitsuitableforshipdetectioninvaryingscenarios. 3.FeatureMatching Inthisstage,wematchtheextractedSIFTkeypointsbetweenadjacentimagestoestablishcorrespondences.TheRANdomSAmpleConsensus(RANSAC)algorithmisadoptedtofilteroutoutliersandobtainaccuratematches.Thematchedkeypointsprovidecriticalinformationforsubsequentshipclassification. 4.ShipClassification Afterfeaturematching,weassignaclasslabeltoeachmatchedkeypointbasedonthepresenceorabsenceofaship.ThisclassificationstageemploysaSupportVectorMachine(SVM)algorithmtrainedonalabeleddataset.TheSVMmodelaccuratelydistinguishesship-relatedkeypointsfrombackgroundkeypoints,enhancingtheoverallshipdetectionperformance. 5.ExperimentalResults Toevaluatetheproposedshipdetectionmethod,weconductedexperimentsonadatasetofon-orbitsatelliteremotesensingimages.Thedataset