基于贝叶斯理论的运动目标跟踪技术分析-analysis of moving object tracking technology based on bayesian theory.docxVIP

AbstractMovingobjecttrackinginvideosequencesisthefundamentalandchallenging researchtopicsintheareaofcomputervisionatpresent.Theyhavegreatresearchand applicationvaluesinmodernindustry,visionnavigation,military,securitymonitoring, intelligenttransportation,videosurveillance,medicaldiagnosisandspaceflightareasetal.Thispaperismainlyconcentratedinthesignalobjecttrackingapplications.Dueto thevideoobjecttrackinghasawiderangeofapplicationinthefieldifcomputervision background,theresearchresultsisalsoofsignificanceforothercomputervision application.Thepaperfortheoutdoortrackfield:toimprovethestabilityoftheobjecttrackingin complexbackground,researchparticlefilterandtheoptimalcolorspaceinPFframework.ParticlefilterrealizerecursiveBayesianfilterviaMonteCarlosimulation.Themethodissuitableforanynon-linearsystemthatcouldberepresentedwithstatemodel.ItismorepracticalthanconventionalKalmanfilteranditsprecisioncouldapproachoptimalestimation.Particlefilterisflexibleandeasytobeimplemented.Anditisalsohasparallelstructure.Themaincontentsinclude:Detailedparticlefiltertheoryanditsimplementation,usingthePFtosolvetheobjecttracking,buildtrackingframeworkbasedonparticlefilter.Forfeatureselection,considerthemovingobjectfeaturedirectimpactontheaccuracyandstabilityofthemovingobjecttracking,theexistingfeaturesneedtoaddress thelargevolumesofdataandcomputationalcomplexity.Sothehistogrammodelhas selectedasafeatureexpressionmanner.Consider the uniformityofthecolorspace,sothe colorspacecan correctly measuretheperceptionofcolorsimilarityandorthogonal.Thispaperstudythreetypesof colorspaceinRGB,HSVandYCbCrfortackingperformanceinfluenceinparticlefilter, toseektheoptimalcolorspaceoftheparticlefilteralgoritm.TheexperimentalresultsshowthattheoptimalcolorspaceinparticlefilteralgorithmisHSVcolor space, which has strong robustness for rigid objects in complex outdoor environment,objectscalechanges,illuminationchangesandpartialocclusion.Keywords:objecttrackingcolorspaceparticlefilteralgorithmevaluation目录引言........