- 1、本文档共8页,可阅读全部内容。
- 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
- 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载。
- 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
IEEEROBOTICSANDAUTOMATIONLETTERS,VOL.8,NO.2,FEBRUARY20231125
NeurAR:NeuralUncertaintyforAutonomous3D
ReconstructionWithImplicitNeuralRepresentations
YunlongRan,JingZeng,ShiboHe,SeniorMember,IEEE,JimingChen,Fellow,IEEE,LinchengLi,
YingfengChen,GimheeLee,andQiYe
Abstract—Implicitneuralrepresentationshaveshowncom-driving,filming,gaming,medicine,architecture.Theproblem
pellingresultsinoffline3Dreconstructionandalsorecentlydemon-requiresarobottomakedecisionsaboutmovingtowardswhich
stratedthepotentialforonlineSLAMsystems.However,applyingviewpointineachsteptogetthebestreconstructionqualityof
themtoautonomous3Dreconstruction,wherearobotisrequiredanunknownscenewiththelowestcost,i.e.viewplanning.In
toexploreasceneandplanaviewpathforthereconstruction,hasthiswork,weassumearobotcanlocalizeitselfandateach
notbeenstudied.Inthispaper,weexploreforthefirsttimethepos-
viewpoint,theinformationofasceneiscapturedbyanRGB
sibilityofusingimplicitneuralrepresentationsforautonomous3D
scenereconstructionbyaddressingtwokeychallenges:1)seekingaimage(withanoptionaldepthimage).
criteriontomeasurethequalityofthecandidateviewpointsfortheImplicitneuralrepresentationsfor3Dobjectshaveshown
viewplanningbasedonthenewrepresentations,and2)learningthetheirpotentialtobepreciseingeometryencoding,efficientin
criterionfromdatathatcangeneralizetodi
文档评论(0)