返回
End-to-end Learning of Multi-sensor 3D Tracking by Detection

End-to-end Learning of Multi-sensor 3D Tracking by Detection

ID：40384361

大小：1.73 MB

页数：8页

时间：2019-08-01

End-to-end Learning of Multi-sensor 3D Tracking by Detection_第1页
End-to-end Learning of Multi-sensor 3D Tracking by Detection_第2页
End-to-end Learning of Multi-sensor 3D Tracking by Detection_第3页
End-to-end Learning of Multi-sensor 3D Tracking by Detection_第4页
End-to-end Learning of Multi-sensor 3D Tracking by Detection_第5页
资源描述：

《End-to-end Learning of Multi-sensor 3D Tracking by Detection》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、2018IEEEInternationalConferenceonRoboticsandAutomation(ICRA)May21-25,2018,Brisbane,AustraliaEnd-to-endLearningofMulti-sensor3DTrackingbyDetectionDaviFrossard1andRaquelUrtasun1Abstract—Inthispaperweproposeanovelapproachtoboththedetectionandtrackingproblems.Wereferthereadertrackingbydetectionthatca

2、nexploitbothcamerasaswellastoFigure1foranoverviewourapproach.LIDARdatatoproduceveryaccurate3Dtrajectories.Towardsthisgoal,weformulatetheproblemasalinearprogramthatII.RELATEDWORKcanbesolvedexactly,andlearnconvolutionalnetworksfordetectionaswellasmatchinginanend-to-endmanner.WeRecentworksinmultiple

3、objecttrackingareusuallydoneevaluateourmodelinthechallengingKITTIdatasetandshowintwofronts:Filteringbasedandbatchbasedmethods.verycompetitiveresults.FilteringbasedmethodsrelyontheMarkovassumptiontoestimatetheposteriordistributionofthetrajectories.I.INTRODUCTIONBayesianorMonteCarlofilteringmethodss

4、uchasGaussianProcesses[3],ParticleFiltersandKalmanFilters[2]areOneofthefundamentaltasksinperceptionsystemsforcommonlyemployed.Oneadvantageoffilteringapproachesautonomousdrivingistobeabletotracktrafficparticipants.istheirefficiency,whichallowsforreal-timeapplications.Thistask,commonlyreferredtoasMult

5、i-targettracking,However,theysufferfromthepropagationofearlyerrors,consistsonidentifyinghowmanyobjectsthereareineachwhicharehardtomitigate.Totacklethisshortcoming,batchframe,aswellaslinktheirtrajectoriesovertime.Despitemethodsutilizeobjecthypothesesfromadetector(trackingmanydecadesofresearch,trac

6、kingisstillanopenproblem.bydetection)overentiresequencestoestimatetrajectories,Challengesincludedealingwithobjecttruncation,highspeedwhichallowsforglobaloptimizationandusageofhighertargets,lightingconditions,sensormotionandcomplexlevelcues.Estimatingtrajectoriesbecomesadataassociationinteractions

7、betweentargets,whichleadstoocclusionandproblem,i.e.,decidingfromthesetofdetectionswhichshouldpathcrossing.belinkedtoformcorrecttrajectories.TheassociationcanbeMostmoderncomputervisionapproachestomulti-targetestimatedwi

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。