Decentralized Particle Filter for Joint Individual-Group Tracking

Decentralized Particle Filter for Joint Individual-Group Tracking

ID:39459244

大小:2.17 MB

页数:8页

时间:2019-07-03

Decentralized Particle Filter for Joint Individual-Group Tracking_第1页
Decentralized Particle Filter for Joint Individual-Group Tracking_第2页
Decentralized Particle Filter for Joint Individual-Group Tracking_第3页
Decentralized Particle Filter for Joint Individual-Group Tracking_第4页
Decentralized Particle Filter for Joint Individual-Group Tracking_第5页
资源描述:

《Decentralized Particle Filter for Joint Individual-Group Tracking》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、DecentralizedParticleFilterforJointIndividual-GroupTrackingLorisBazzania;bMarcoCristania;bVittorioMurinoa;baUniversityofVerona,StradaleGrazie,Verona,ItalybPatternAnalysis&ComputerVision,IstitutoItalianodiTecnologia,ViaMorego,Genova,Italyname.surname@itt.itAbstractInthispaper,weaddressthet

2、askoftrackinggroupsofpeopleinsurveillancescenarios.Thisisamajorchallengeincomputervision,sincegroupsarestructuredentities,subjectedtorepeatedsplitandmergeevents.Oursolutionisajointindividual-grouptrackingframework,inspiredbyarecenttechniquedubbeddecentralizedparticlefiltering.Theproposedst

3、rategyfactorizesthejointindividual-groupstatespaceintwodependentsubspaceswhereindividualsandgroupssharetheknowledgeofthejointindividual-Figure1:Grouptrackingresults(coloredconvexhulls)ofgroupdistribution.Inpractice,weestablishatightrelationtheproposedFriendsmeetdataset(firstrow)andofBIWIof

4、mutualsupportbetweenthemodelingofindividualsanddataset(secondrow)[22].thatofgroups,promotingtheideathatgroupsarebettertrackedifindividualsareconsidered,andviceversa.Exten-groupstatespaceintwoconditionallydependentsubspaces,siveexperimentsonapublishedandnoveldatasetvalidatesothatitispossib

5、letomodel1)thesingleindividuals,andourintuition,openinguptomanyfuturedevelopments.2)thegroupsgiventheknowledgeoftheindividuals.WedubbedourproposalasDEEPER-JIGT:DEcentralizEdPar-ticlefiltERforJointIndividual-GroupTracking.1.IntroductionManyinterestingqualitiescanbeascribedtoDEEPER-Grouptrac

6、kingconsistsinfollowingtightformationsofJIGT.Firstofall,theabsenceofheuristicstohandlegroupindividualswhiletheyarewalkingorinteracting(Fig.1).events:theyareallgovernedbyprobabilitydistributionsThisrecentopenchallengeisimportantinmanyrespects:inwhoseparameterscanbelearnedfromtrainingdata.M

7、ore-computervisionandsignalprocessing,itmayhelpinlocat-over,DEEPER-JIGTupdatesthegroupinformationinaningindividualtargetsinthecaseofmissingmeasurementsonlinefashion,wherethetrackinghistoryoftheindividuals[21,22];insurveillance,itmayrevealsocialbondsbetweenisintrinsi

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。