资源描述：
《基于视觉适应的在线目标追踪算法分析与实现》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、摘要一视频帧，目标的观测模型都是通过联合检测器与在线的分类器构造得到。综上所述，本文详细分析与实现了三种基于不同类型的在线目标追踪算法。同时，我们在不同的数据库上进行实验验证以及与其它方法进行比较。实验结果表明可以将实现大量图片或者非当前视频帧的视觉知识迁移到在线的目标追踪，这种视觉知识的迁移是有益的。关键词：目标追踪，域适应，在线学习，分类器，视觉字典IIABSTRACTABSTRACTObjecttrackingisanimportantresearchblanchandbasicprobleminthefieldsofcomputervision.Meanwhile,it’s
2、alsotheimportantfoundationofmanyhigh-levelapplications,whicharebasedontheanalysisofobjecttracking,suchasbehaviorrecognition,behavioranalysis,etc.Butalsoforintelligentmonitoring,human-computerinteraction,transportationmanagement,medicalimageprocessing,etc.providebasicdatafields,sotheissuein-dep
3、thstudyhasimportantsignificanceandpracticalvalue.Learningtoadapttotheuseofthedomaintheorytosolvethetargettrackingproblemincomputervision,becauseitcanusethelineofvisionofknowledgeandalotofpicturessothattheprocessofupdatingtheclassifiersingleusesampledataisnolongercurrentframe,awiderangeofdomest
4、icandforeignscholarsinrecentyearsbyattention.Thisthesisisbasedontheanalysisoftraditionalobjecttrackingalgorithm,andfocusedononlineobjecttrackingbasedondomainadaptionlearningtheory.Theconcreteworkisasfollow:1.Wedescribedandanalyzedthebasicclassicalcorepartoftheonlineobjecttrackingalgorithm,name
5、ly,trainedaclassifiertodistinguishthebackgroundandtheforeground,includingP-NlearningtheoryofTLDobjecttrackingalgorithm,multipleinstancelearningofMILtracingalgorithmandtheonlineboostingalgorithm.Throughstudyofthecoreoftheonlinetargettrackingalgorithmfounditsshortcomings,combinedwiththeactualsit
6、uationofthetargettrackingbasedonnecessityandfeasibilityofproposeddomainadaptationlearningobjectivestotrack.2.Weputforwardtheadaptiveonlinetargettrackingalgorithmbasedontheinstance.Inthispaper,thetargettrackingonlinetraininglearningclassifierasdomainadaptivelearningproblems,previousimagesequenc
7、etrackingtargetinstanceknowledgetransfertothefirstt-1frameimage.Accordingtotheseinstancesofknowledgewiththefirstt-1frameoftargetinformationonlinetrainingaclassifier,andthemethodofusingtracking-checktotargetthefirsttframe.3.Weofferonline
