基于顏色和頭部輪廓的粒子濾波跟蹤算法.pdf

1、中山大學(xué)碩士學(xué)位論文基于顏色和頭部輪廓的粒子濾波跟蹤算法姓名：孫國偉申請學(xué)位級別：碩士專業(yè)：應(yīng)用數(shù)學(xué)指導(dǎo)教師：戴道清20060528中山大學(xué)碩士學(xué)位論文：基于顏色和頭部輪廓的粒子濾波跟蹤算法Head—trackingBasedonColorandContourwithParticleFilterMajor：AppliedMathematicsName：GnoweiSunSupervisor：ProfessorDao—qingDaiAbs

2、tractMovingobjectstrackingisoneoftheactiveresearchdirectioninthefieldofcomputervisionPeoplegettheexotericinformationbysense，andvisioninformationisthemainknowledgetheygetImagesequencesofmovingobjectcontainmoreinformationt

3、hanstaticimages，whileobjecttrackingisanewresearchsubjectwiththedevelopmentofdigitalvideofrequencytechnologyIthasimportantsignificanceinscienceandengineeringandpotentiallywidespreadapplicationsinmanydomains，suchassmartsur

4、veillance，perceptualinterface，motionanalysis，virtualrealityInthispaper，weresearchtheheadtrackingofsinglepeopleinimagessequencesThethesisdiscussesthegeneralobjectstrackingproblemundertheframeofBayesianestimationandthecomm

5、onalgorithmofparticlefilterTheparticlefilterbasedonSequentialImportanceSamplingisdeducedindetailWepresentaparticlefilterbasedvisualtrackingwithcolorandshapecuesfusionVarioustrackingalgorithmshavebeenpropose，however，mosto

6、fthemarebasedonasingleimagecueItisclearthatmultiplecuesfusionCanprovidemorereliableobservationInthispaper，weuseanellipsetepresentthehumanheadCombiningtheknowledgeofheadcontourandfacecharacteristic，weinosculatetwodifferen

7、tmodelstepresentOurobject，oneisellipsemodelbasedoncontourerror，theotheriscolorhistogrammodelbasedoncolorinformationofinnerobjectareaThenononehandwecanimprovethecorrectnessbypreciseorientationofedge，ontheotherhand，improve

8、robustnessofalgorithmbyrichinformationofinnerobjectWhenconsideringthespacecontinuityandregularityofthemovingobjectbetweenneighborhoodframes，weuseadaptivevelocitymodeltodescribemovingobject，thenuseparticlefiltertotrackthe