动态参数估计

动态参数估计

ID:39256717

大小:519.82 KB

页数:4页

时间:2019-06-28

动态参数估计_第1页
动态参数估计_第2页
动态参数估计_第3页
动态参数估计_第4页
资源描述:

《动态参数估计》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、Multi-ViewActiveShapeModelwithRobustParameterEstimationLiZhang,HaizhouAiDepartmentofComputerScienceandTechnology,TsinghuaUniversity,Beijing,100084,Chinaahz@mail.tsinghua.edu.cnAbstracttolocalizeobjectswithlargeviewchanges,especiallyfortheapplicationofmulti-viewfacealignment.TheseworksActi

2、veShapeModelisanefficientwayforlocalizingob-ofteninvolvemultipleshapeandtexturemodelsbuiltbyjectswithvariableshapes.WhenASMisextendedtomulti-non-linearmethodssuchasGaussianmixtures[1]andKer-viewcases,theparameterestimationapproachesinprevi-nelPCA[6],whileself-occlusioncanalsobehandledbyous

3、worksareoftensensitivetotheinitialview,astheydolearningvisibilityfromtrainingdata[11].However,sincenothandletheunreliabilityoflocaltexturesearch,whichthetexturemodelusedinthelocalsearchofeachlabelpointcanbecausedbybadinitializationorclutteredbackground.dependsontheview,thesemethodsareofte

4、nverysensi-Toovercomethisproblem,weproposeanovelalgorithmtivetotheestimationofthehiddenviewparameter.Whenforparameterestimation,usingrobustestimatorstoremovetheinitialviewisnotpredictedcorrect,theresultsoflo-outliers.Byweightingdynamically,ourmethodactsasacalsearchbecomeunreliable.Ifthees

5、timationoftheshapemodelselectionmethod,whichrevealsthehiddenshapeparameterdoesnotdealwiththepotentialoutliers,theseandviewparametersfromnoisyobservationsoflocaltex-multi-viewASMapproacheswillfail.turemodels.Experimentsandcomparisonsonmulti-viewToovercomethisproblem,weproposeanovelparame-f

6、acealignmentarecarriedouttoshowtheefficiencyofourterestimationapproachformulti-viewActiveShapeModel.approach.Inourmethod,everylabelpointisweighteddynamically.Onlythelabelpointsthatareconsistentwiththeshapemodelwillhavelargeweights,whileoutliershavelittle1.Introductionweightsandtheirinfluenc

7、eareeliminated.Bydoingthis,thealgorithmactsasamodelselectionmethodrevealingthehiddenviewparameterfromunreliableobservationsofActiveShapeModel(ASM)introducedbyCootes[2]labelpoints.Sinceourmethoddoesnotcompletelydependisapopularmethodforobjectregistrationinimages.Byon

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。