Visual Tracking

Visual Tracking

ID:39195110

大小:191.92 KB

页数:16页

时间:2019-06-26

Visual Tracking_第1页
Visual Tracking_第2页
Visual Tracking_第3页
Visual Tracking_第4页
Visual Tracking_第5页
资源描述:

《Visual Tracking》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库

1、ECE432-AdvancedComputerVisionNotesSeries7VisualTrackingYingWuElectrical&ComputerEngineeringNorthwesternUniversityEvanston,IL60208yingwu@ece.nwu.eduContents1GettingStart–BackgroundSubtraction22TemplateMatching32.1SSDandCross-Correlation...............

2、.............32.2Searchvs.Descent................................32.2.1Search...................................32.2.2TheDirectMethod............................43TwoPhilosophies53.1Bottom-upApproach...............................53.2Top-downApproach

3、...............................54AProbabilisticFramework64.1WhatisTracking?................................64.2FourElementsinTracking............................64.3AnexampleofObservationModel........................84.4TrackingasProbabilityPropagati

4、on......................85KalmanFiltering95.1KalmanFilteringforLinearDynamicSystems.................95.2Example:TrackingFeaturePoint........................106SequentialMonteCarlo126.1FactoredSampling................................126.2ImportanceSam

5、pling...............................1311GettingStart–BackgroundSubtractionInmanyvideosurveillanceapplications,camerasarefixedandweareinterestedintrackingthemotionoftheforeground,whichcouldbepeopleorcars.Obviously,framedifferenceonlygivesusaroughideaofw

6、hichregionsmaycontainmovingobjects,butsuchasimplemethodcanneithersperatetheforegroundfromthebackground,nortelluswhichimageregionsaremovingregions.Asaresult,doingsimpleframedifferencingwillnotoutputgoodtrackingresults.Backgroundsubtractionissuchatechni

7、quethattheforegroundregioncanbeseparatedfrombackgroundbymaintainingabackgroundmodel,andclassifyingeachpixelintoeitherforegroundandbackground.Theassumptionhereisthatthecameraismoreorlessfixedsuchthatwecanmaintainortrainabackgroundmodel.Generally,suchba

8、ckgroundmodelcanbepixel-wise,i.e.,eachpixelismodeledindependently.Foreachimagepixellocation,wecomparetheinputimagepixelanditsbackgroundmodelanddetermineitsclass.Asimpleapproachjustusesameanimageasthebackground,thosepixelswhichdeviatefromthecorrespond

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。