预览加载中，请您耐心等待几秒...

基于点云分割的运动目标跟踪与SLAM方法.docx

预览
1/3
2/3
3/3

在线预览结束，喜欢就下载吧，查找使用更方便

下载提示 文本预览

如果您无法下载资料，请参考说明：

1、部分资料下载需要金币，请确保您的账户上有足够的金币

2、已购买过的文档，再次下载不重复扣费

3、资料包下载后请先用软件解压，在使用对应软件打开

基于点云分割的运动目标跟踪与SLAM方法 Abstract Objectdetection,tracking,andlocalizationarecrucialpartsoftheautonomousnavigationsystem.Thesetasksareparticularlychallengingwhenitcomestooutdoorenvironmentsthatareoftenclutteredandhavedifferentkindsofobjectsmovingindifferentdirections.Inrecentyears,advancesinlidartechnologyhaveallowedforrobustandaccurateobjectdetectionandlocalizationusingpointclouddata.Inthispaper,weproposeamethodformotionobjecttrackingandsimultaneouslocalizationandmappingusingpointcloudsegmentation.Theproposedmethodisbasedondeeplearning-basedpointcloudsegmentationandrandomsampleconsensus(RANSAC)algorithmfor3Dobjectdetection,tracking,andlocalization. Introduction Autonomousnavigationinoutdoorenvironmentsisincreasinglyimportantasvehiclesbecomeincreasinglyautomated.Forautonomousvehiclestobesuccessfulintheseenvironments,theymustbeabletoperceiveandreacttotheirsurroundingsquicklyandaccurately.Objectdetection,tracking,andlocalizationarekeyelementsoftheautonomousnavigationsystem.Detectingandtrackingmovingobjects,suchaspedestrians,isparticularlychallengingbecauseoftheconstantchangesintheenvironment. Recentstudieshaveshownthatlidar-basedsensorscanbeusedtodetectandtrackobjectsin3Dspace.Lidarsensorsgeneratepointclouddatathatallowsforaccurateobjectlocationandclassification.Pointcloudsegmentationistheprocessofdividingapointcloudintodifferentcomponentsorpartsthatcanbeusedforobjectdetectionandlocalization. Inthispaper,weproposeamethodformotionobjecttrackingandsimultaneouslocalizationandmapping(SLAM)usingpointcloudsegmentation.Theproposedmethodcombinesdeeplearning-basedpointcloudsegmentationandrandomsampleconsensus(RANSAC)algorithmfor3Dobjectdetection,tracking,andlocalization. PointCloudSegmentation Pointcloudsegmentationistheprocessofdividingapointcloudintodifferentcomponentsorpartsthatcanbeusedforobjectdetectionandlocalization.Variousmethodshavebeenproposedforpointcloudsegmentation,includingsupervisedandunsupervisedlearningtechniques. Supervisedlearningtechniquesrequirealabeleddataset