基于多特征融合的室外车辆鲁棒跟踪分析-robust tracking analysis of outdoor vehicles based on multi-feature fusion.docxVIP

AbstractSupervisor:Signature:AbstractThevehicletrac挝ngtechnologyisthecorecontentandk巳ytechnologyinintelligenttransportati册，andisthepreconditionofbehavio玄analysisandidentificationofthevehicle.Intra丑Eicvideosurveillance，detectionandtrackingc皿bensedtogetinformationandoverallplanningtransportation.WiththecontinuedgrowthintrafEicvolume，位afEcicongestionand悦。ckin辜缸egrowing，trafEicpollutionandtrafEicaccident加vecausedwidespread∞ncemofthe胃holesocie纱.Theuseofinte以革ent饺ansportationsystems(IτS)toimprovetheutilizationoftheroad，tr必1icsa句句roftheroadandt挝eroad例邵lllg虹肝ebecomethe知他血directionoftra忍sportation.Thevehicletracking忌as忱enmademanyefficienlaJgorithms，butt汝restillhavemanydifficnltiesinpracticalapplicatio旦sofvideoobjecttracki且在，such部也eillnminationchanges，theposechang臼ofthetracki吨target，t泌interl泣咄咄船dnoiseofbackgronnd.How10a乙hieve也erobustvehicletrackingintheoutdoorisstillachallengingtaskInthispaper，we皿mat也edi部cnltproblemsofoutdoorvehicletrackingtechnology，targettrackingalgori钱mwl让你isbasedonadaptivemnlti-f切盐rrefusionisproposedund町也eframe听orkofparticJefilte丑Thispaperfir百tin甘oduceshowtomeasure也eunce此必ntyof社1号feato耐烦恼gettracking，whichcanuse世ledistribulionofeac且partícleand也elikelihoodfunctiontome那时ethere挂abilítyof础chfeature.Under也econstantcondítionsofthetotaln田怒.berofparticles，fea沁re最lsionshould.be部medat岱C照丑必ilityofeachfea:也胆，deter双ínethe口.umberoft如eoptímaJp盯ticleselected企o皿eachf社ature.了。.beabletoadaptivelyac号ust也ewel趴tsofeachfeaturε，也ispaper船alyzesthedístri七四onof也eparticlesanddesi在nedthew白ghtsadaptiveupdatealgorithm.Thismet险。dbased。但也ecnrrentframepartìcledís位ibutioncanwellweí约thereliabili守ofeachfeature，and10adjust如weightsaccordingly.Experimentalresnltsshow鼠就命eproposedaJgorithmcomparedwi吻。thermethodsindífferenttrac溢血gconditionsachievedsi革nificantimprovementsinresnlts.Keyword锺:targettrackin莓;particle丑lter;照nltí-必at四e盐lsion;adap缸veweíghts目录1绪论11.1研究背景及意义11.2国内外研究现状21.2.1目标跟踪方法21.2.2基于多特征融合的目标跟踪算法研究现状61.3本文主要内容及章节安排72基于粒子滤波的目标跟踪92.1引言92.2粒子滤波原理92.2.1蒙特卡洛方法10贝叶斯重要性采样（BayesianImportanceSamp