视频的车辆检测与跟踪算法综述

　29　2()Vol.29　No.2　　20094JournalofNanjingUniversityofPostsandTelecommunications(NaturalScience)Apr.2009董春利1,2,董育宁1,31.,　2100032.,　2111883.,　215006　:首先介绍了交通检测系统,指出视频交通检测技术日益成为计算机视觉领域中备受关注的前沿方向。在此基础上,分别讨论了常用的车辆检测算法,基于模型的车辆检测算法,车辆跟踪的基本类型,以及基于模板匹配、卡尔曼滤波和粒子滤波的车辆跟踪算法,同时分析比较了各种算法的优缺点。最后,展望了这一领域未来研究的热点。:智能交通系统;交通检测系统;车辆检测;车辆跟踪:TP391.41　　　:A　　　:1673-5439(2009)02-0088-07SurveyonVideoBasedVehicleDetectionandTrackingAlgorithmsDONGChun-li1,2,DONGYu-ning1,31.CollegeofTelecommunicationsInformationEngineering,NanjingUniversityofPostsandTelecommunications,Nanjing210003,China2.DepartmentofInformationEngineering,NanjingCollegeofTransportationVocationalTechnology,Nanjing211188,China3.JiangsuProvincialKeyLabofComputerInformationProcessingTechnology,SoochowUniversity,Suzhou215006,ChinaAbstract:Thetrafficdetectionsystemisintroducedfirst,anditispointedoutthatthevideobasedtrafficdetectiontechniqueisincreasinglybecomingthemostactiveresearchtopicsinthefieldofcomputervi-sion.Thenvariousalgorithmsarediscussedrespectively,suchasthecommonlyusedvehicledetectional-gorithms,themodelbasedvehicledetectionalgorithms,thebasictypesofvehicletracking,aswellasthetemplatematching,KalmanfilteringandParticlefilteringbasedvehicletrackingalgorithms.Theadvanta-gesanddisadvantagesofvariousalgorithmsarecomparedindetail.Finally,thehotspotsforfuturere-searchinthisareaarepresented.Keywords:intelligenttransportsystems;transportationdetectionsystems;vehicledetection;vehicletrack-ing:2008-10-11:(205060)、(07KJA51006)、(KJS0712)、:　:[phone]　E-mail:[e-mail]0　　(IntelligentTransportSystems,ITS)、、、、,、、[1-2]。[3-4]60,,。20014,260,10。,,ITS、、。、ITS,,ITS,,。,ITS600,,。,ITS,ITS,、。。,、、,。,,[5]。“”,“”,ITS、、、,2008、2010。[6],,,1。1　,、,、,、、、,,,,,,。。1　,,。。、、、,,。。1.1　1.1.1　[7],。,,;,。,,。;,,;,。,[8],。1.1.2　Gibson1950。,(2D),。[9],。,89　2:,()。,,、、;,,,。1.1.3　[10],,。,,。[11]。,。,。[12]。。,,[13]。,,,。、。。,。[14],,。1.1.4　,,。(),,,,。[15]3。[16],。,,。,,。,,。,,,。1.2　1.2.1　3-D3-D3-D,。3-D:(),。[17],3-D。,12,,,12。,,。[18],。,。,,,。3-D。、、,3-D,。,3-D,(、,),。1.2.2　,[14]。90()　　　　　　　　　　2009　。,(、)K。K,。,,;(),。:、;,,,,。2　,、,。,,。,Hausdorff。,。,、、。、、,:、,。,。2.1　2.1.1　[19-20]。,3-D。。,。,,,。2.1.2,,,。[21],、。,。2.1.3　,[22]。,。,,。,。,snake,。,。,。2.1.4　,[23]。,。、,、、、,。,,。,。,。2.2　2.2.1　[24-25](),。,91　2:。,,、。T,M×N,S,L×W。TS,Si,j,i,jS,。i,j:1≤i≤L-M+1,1≤j≤W-N+1,TSi,j,,TSi,j。TSi,j。　D(i,j)=∑Mm=1∑Nn=1[Si,j(m,n)-T(m,n)]2(1)(1),:R(i,j)=∑Mm=1∑Nn=1[Si,j(m,n)-T(m,n)]2∑Mm=1∑Nn=1[Si,j(m,n)]21/2∑Mm=1∑Nn=1[T(m,n)]21/2(2),(2)0≤R(i,j)≤1,R(i,j),Si,jT,。,。[25],,,,。2.2.2　[26],,,(),,。,,。,,。:,,。k-1p(xk-i/zk-1),:　　Xk=k,k-1Xk-1+Bk-1Uk-1+Γk-1Wk-1Yk=HkXk+Vk(3),Xk-1k-1;kk-1k;Bkk-1;Ukk-1;Wk-1k-1,;Γk-1;Yk;Hkk;Vkk。WkVk。,::Xk/k-1=k,k-1Xk-1+Bk-1Uk-1(4)Xk/k-1Xk,Xk-1,Xk-1k-1。:Xk=Xk/k-1+Kk(Yk-HkXk/k-1)(5)XkYkXk/k-1,k。[26-27],,,;,,。,,;、,。。[28],GM(1,1)。GM(1,1),,,,,。2.2.3　,,,。,92()　　　　　　　　　　2009　z1:kxkp(xkz1:k),。[29],。,,,,。,(BIS),q(xi0,k/z1:k),p(xi0,k/z1:k)。,BIS,SIS。:ωik=ωik-1p(zk/xik)p(xik/xik-1)q(xik/xik-1,z1:k)(6):q(xik/xik-1,z1:k)=p(xik/xik-1)(7)(6):ωik=ωik-1p(zk/xik)(8)SIS,,Ns,(xik)Nsi=1,,,{xi0:k,N-1},,:E(g(xi0:k))=1Ns∑Nsi=1g(xi0:k)(9),,,,。[30],。[31],,。,。,。,。3　,。、、,,,。,:,,、;,。:、、;,。,。,。。。、、,、。:[1]TOMIZUKAM.AnIntelligentTransportationSystemfortheNextCentury[J].IEEEInternationalSymposiumonIndustrialElectron-ics,1997(1):1-4.[2].ITS[J].,2004(2):17-24.[3]JIXiaopeng,WEIZhiqiang.EffectiveVehicleDetectionTechniqueforTrafficSurveillanceSystems[J].JournalofVisualCommunica-tionandImageRepresentation,2006(17):647-658.[4].[EB/OL].http:∥[5],.2004[J].,2004(2):10-15.[6]WANGYK,CHENSH.RobustVehicleDetectionApproach[C]∥ProcIEEEConferenceonAdvancedVideoandSignalBasedSur-veillanceComo.Piscataway:IEEE,2005:117-122.[7]SEKIM,FUJIWARAH,SUMIK.Arobustbackgroundsubtraction93　2:methodforchangingbackground[C]∥ProceedingofIEEEWork-shoponApplicationsofComputerVision.Piscataway:IEEE,2000:207-213.[8]COLLINSR,LIPTONA,KANADET.AsystemforVideoSurveil-lanceandMonitoring[C]∥Procof8thInternationalTopicalMeet-ingonRoboticsandRemoteSystems.Pittsburgh:ANS,1999:1-68.[9]HARALICKRM,LEEJS.TheFacetApproachtoOpticalFlow[C]∥ImageUnderstatingWorkshop.Arlington:[s.n.],1984:74-83.[10]HORPRASERTT,HARWOODD,DAVISLS.AStatisticalAp-proachforReal-timeRobustBackgroundSubtractionandShadowDetection[C]∥ProceedingsofIEEEICCV'99Frame-RateWorkshop.Kerkyra:IEEE,1999.[11]ABUTALEBAS.AutomaticThreshholdingofGraylevelPicturesUsingTwo-dimensionalEntropy[J].ComputerVisionGraphicsandImageProcessing,1989,47(2):22-32.[12]PARKY.ShaperesolvingLocalThreshholdingforObjectDetection[J].PatternRecognitionLetters,2001,22(5):883-890.[13]BLOSSEVILLEJM,KRAFFTC,LENOISF,etal.TITAN:NewTrafficMeasurementsbyImageProcessing[C]∥IFACSymposiumCCCT'89.Patis:PergamonPress,1989.[14]STAUFFERC,GRIMSONW.AdaptiveBackgroundMixtureMod-elsforReal-timeTracking[C]∥ProceedingsoftheIEEEConfer-enceonComputerVisionandPatternRecognition.Piscataway:IEEE,1999:246-252.[15]ABRAMCZUKT.AMicrocomputerbasedTVDetectorforRoadTraffic[J].SymposiumonRoadResearchProgramT,1984,3(2):145-147.[16],,.[J].,2002,23(4):386-391.[17]KOLLERYD,DANIILIDISY