visual object tracking based on mean-shift and particle-kalman filter视觉物体跟踪基于均值漂移和particle-kalman过滤器

《visual object tracking based on mean-shift and particle-kalman filter视觉物体跟踪基于均值漂移和particle-kalman过滤器》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、硕士留学生学位论文VisualObjectTrackingBasedonMean-shiftandParticleandandPPartP-KalmanFilter作者姓名IreneAnindaputriIswanto学科专业ElectricalandComputerEngineering指导教师李彬副教授所在学院AutomationScienceAndEngineering论文提交日期2016年5月23日VisualObjectTrackingBasedonMean-shiftandParticle-KalmanFilterADissertationSubmi

2、ttedfortheDegreeofMasterCandidate：IreneAnindaputriIswantoSupervisor：AssociateProfessorBinLiSouthChinaUniversityofTechnologyGuangzhou,China分类号：学校代号：10561学号：201422800063华南理工大学硕士学位论文VisualObjectTrackingBasedonMean-shiftandParticle-KalmanFilter作者姓名：IreneAnindaputriIswanto指导教师姓名、职称：李彬，副教授

3、申请学位级别：工程硕士学科专业名称：电气与计算机程研究方向：图像处理论文提交日期：2016年05月23日论文答辩日期：2016年06月08日学位授予单位：华南理工大学学位授予日期：2016年06月25日答辩委员会成员：主席：LUGUONENGG委员：裴海龙，苏为洲，李向阳，李彬gABSTRACTDuetotheincreasingofvideosurveillancesystemrequirements,Intelligentvideosurveillancesystemhasbecomechallengingtopicincomputervisionresea

4、rchfield.Therearefourkeystepsinintelligentvideosurveillancesystem,i.e.objectdetection,objectclassification,objecttracking,andobjectanalysis.Amongthesesteps,objecttrackingisconsideredascrucialandsignificanttaskinintelligentvideosurveillancesystem.Objecttrackingisconsideredasdifficultt

5、askbecauseofseveralproblemssuchasilluminationvariation,trackingnon-rigidobject,non-linearmotion,occlusion,andrequirementofrealtimeimplementation.Thereforeitisnecessarytobuildavisualobjecttrackingalgorithmwhichcanovercometheseproblems.Everysinglealgorithminvisualobjecttrackingalwaysha

6、sbothstrengthsanddrawbacks.Therefore,utilizingonlyonesinglealgorithmfortrackingusuallyisconsideredasinefficientbecauseeverysinglealgorithmhaslimitations.Basedonthisreason,inthisthesisatrackingalgorithmwhichcombinesmean-shiftandparticle-Kalmanfilterisproposed.Intheproposedmethod,mean-

7、shiftisusedasmastertrackerwhenthetargetobjectisnotoccluded.Whenocclusionisoccurredorthemean-shifttrackingresultisnotconvincing,particle-Kalmanfilterwillactasmastertrackertoimprovethetrackingresults.Experimentalshowsthattheproposedmethodcanworkwellindealingwithtrackingproblemssuchasno

8、n-rigidobjec