搜索
365201109ScienceofSurveyingandMappingVol.36No.5Sep.1983-。E-mailnunu328@163.com2010-04-07RANSAC①②③④③①100101②96633100096③61512100088④450052【】。Ran-domSamplingConsensusr。【】r【】TP751【】A【】1009-2307201105-0144-031、、1。2。345。。。2RANSACRANSACRandomSamplingCon-sensusFishierBolles6。。MPP＞εP=1－1－1－εmM1εm。RANSAC、、7。BauerRANSAC8。RANSAC。3RANSAC3.1。ax+by+cz=d2xyzabca2+b2+c2=1d。。。xiyiz{}ii=12…nnFxiyizi－[]1F=03F=[]abcdT2F33。3。。PxyzPLabcddPPL=ax+by+cz－d4δ°。δ°。δ°δ°δ°。δ°。1。5RANSAC。2。2r。。rr2rnumrnumr°。3.2。①②。3L1、L2p1、p2n→1、n→2p1、p2n1→＇//n→1p1p2r→12。①θθ=cos－1n→1·n→26②dΔd=|d1－d2|Δd。θΔd。Δd=max|r→12·n→1||r→12·n→2|7θ°d°。33.3RANSACRANSAC。。SnumSnum。RANSAC1M。21mm=3。34δδ°。42～3M。5r。Snum。62～6。7①②。4LandMark60～80pt/m2。4a、、。δ°=0.08mr°=1mrnum=10。51ε=0.299%M1000。4。。θ°=0.1Δd°=0.5。4c4b。20.39s40.11s。42。1ABA∩BABA/BABB/ABA。1ABA∩BA/BB/A1124601332412310150101422528262324992912411212310249998.80%98.85%AB15029r。51RANSAC13854136。。AIRSARLSAR。XSARP。SAR。1LeeJSGrunesMRGrandiGDe.PolarimetricSARspecklefilteringanditsimplicationforclassificationC//IEEETrans.onGeosci.RemoteSensing19993752363-2373.2LeeJS.SpeckleAnalysisandSmoothingofSyntheticApertureRadarImagesJ.ComputerGraphicandIm-ageProcessing198117324-32.3NovakLMBurlMC.Optimalspecklereductioninpo-larimetricSARimageryC//IEEETrans.Aerosp.Electron.Syst1990262293-305.4LeeJSGrunesMRMangoSA.SpecklereductioninmultipolarizationmultifrequencySARimageryC//IEEETransactionsonGeoscienceandRemoteSensing1991294535-544.5GozeSLopesA.AMMSEspecklefilterforfullresolu-tionSARpolarimetricdataJ.JournalofElectromag-naticWavesandApplications199375717-737.6.M.2008.7LeeJSGrunesMRetal.Scattering-model-basedspecklefilteringofpolarimetricSARdataC//IEEETransactionsonGeoscienceandRemoteSensing2006441176-187.8YoshioYamaguchiToshifumiMoriyamaetal.Four-ComponentScatteringModelforPolarimetricSARImageDecompositionC//IEEETransGeosciRemoteSens-ing20054381699-1706.9LeeJS.DigitalImageEnhancementandNoiseFilteringbyuseofLocalStatisticsC//IEEETransactionsonPatternAnalysisandMachineIntelligence1980PA-MI-22165-168.10.SARJ.2008133377-385.AnimprovedmethodofLeerefinedpolarizedfilterAbstractAnimprovedmethodforovercomingthedeficiencyofLeerefinedpolarizedfilterinpreservingpolarizedinformationwasproposedinpresentstudy.Toimplementthismethodthepaperfirstlydividedallpixelsintofourscatteringcategoriesbasedonthefour-componentdecompositionalgorithm.Thenpixelswithdifferentscatteringcharacteristicstothecentralizedonewereremovedinedge-alignedwindow.Finallyspecklefilteringwasutilizedaccordingtothelocalstatisticalpropertiesofhomogeneousare-a.Experimentalresultsshowedthattheproposedmethodcouldperformbetterinpreservingpolarizedinformationthantraditionalone.KeywordspolarizedSyntheticApertureRadarspeckleLeerefinedpolarizedfilterfour-componentscatteringmodelOUYANGQun-dongWUZhao-congPENGJian-guiSchoolofRemoteSensingandInformationEngineeringWuhanUniversi-tyWuhan430079China檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿檿145。2r。3r。4。1.M.2003.2.J.200232295-100.3WaniMAArabniaHR.Paralleledgeregion-basedsegmentationalgorithmtargetedatreconfi-gurablemulti-ringnetworkJ.JournalofSupercomputing.200325143-62.4SitholeGVosselmanG.Automaticstructuredetectioninapoint-cloudofanurbanlandscapeC//2ndJointWorkshoPonRemoteSensingandDataFusionoverUr-banAereasUrban2003.5RabbaniTHeuvelFAVosselman.SEGMEN-TATIONOFPO-INTCLOUDSUSINGSMOOTH-NESSCONST-RAI-NTJ.*inIAPRSVolumeXXXVIPart5.Dresden.2006.6FischlerMABollesRC.RandomSam-pleConsensusAParadigmforModelfittingwithap-plicationtoImageAnalysisandAutomatedCartogra-phyJ.Communica-tionsoftheACM1981246381-395.7.J.20051681431-1438.8BauerJKarnerkSchindlerKKlausAZachC.Segmentationofbuildingfromde-nse3Dpoint-clouds.inProceedingsoftheISPRSM.WorkshoPLaserscanning.2005.EnschedetheNetherlands.SegmentationofbuildingfacadepointcloudsusingRANSACAbstractSegmentationofbuildingfacadepointcloudsisthefoundationoffeatureextractionandmodelingfromVehicle-BorneLi-DAR.InthepaperRandomSamplingConsensuswasintroducedintothesegmentationofLiDARandr-radiuspointdensitywasputfor-wardtotheestimationcriterionwhichaimstoremovethediscretepointoutsidethefeatureplane.Thentwoconstraintsofangleanddistancewereerectedtounitethesegmentedplaneswhichoptimizedtheresults.Keywordsvehicle-borneLiDARRANSACsegmentationofpointcloudsr-radiuspointdensityLINa①②MAYi-wei③YANGYang④GAOSheng-li③①InstituteofRemoteSensingApplicationsChineseAcademyofSci-encesBeijing100101China②Troops96633Beijing100096China③Troops61512Beijing100088China④InstituteofSurveyingandMappingInformationEngineeringUniversityZhengzhou450052China831