The three-dimensional motion of humans is underdetermined when the observation is limited to a single camera, due to the inherent 3D ambiguity of 2D video. We present a system that reconstructs the 3D motion of human subjects from single-camera video, rely
BayesianReconstructionof3DHumanMotionfromSingle-CameraVideo
NicholasR.Howe
DepartmentofComputerScience
CornellUniversityIthaca,NY14850nihowe@cs.cornell.eduMichaelE.LeventonArti cialIntelligenceLab
MassachusettsInstituteofTechnology
Cambridge,MA02139leventon@ai.mit.edu
WilliamT.Freeman
MERL–aMitsubishiElectricResearchLab
201BroadwayCambridge,MA02139freeman@merl.com
Abstract
Thethree-dimensionalmotionofhumansisunderdeterminedwhentheobservationislimitedtoasinglecamera,duetotheinherent3Dambiguityof2Dvideo.Wepresentasystemthatreconstructsthe3Dmotionofhumansubjectsfromsingle-cameravideo,relyingonpriorknowledgeabouthumanmotion,learnedfromtrainingdata,toresolvethoseambiguities.Afterinitializationin2D,thetrackingand3Dreconstructionisautomatic;weshowresultsforseveralvideosequences.Theresultsshowthepoweroftreating3Dbodytrackingasaninferenceproblem.
1Introduction
Weseektocapturethe3Dmotionsofhumansfromvideosequences.Thepoten-tialapplicationsarebroad,includingindustrialcomputergraphics,virtualreality,andimprovedhuman-computerinteraction.Recentresearchattentionhasfocusedonunencumberedtrackingtechniquesthatdon’trequireattachingmarkerstothesubject’sbody[4,5],see[12]forasurvey.Typically,thesemethodsrequiresimul-taneousviewsfrommultiplecameras.
Motioncapturefromasinglecameraisimportantforseveralreasons.First,thoughunderdetermined,itisaproblempeoplecansolveeasily,asanyoneviewingadancerinamoviecancon rm.Singlecamerashotsarethemostconvenienttoobtain,and,ofcourse,applytotheworld’s lmandvideoarchives.Itisanappealingcomputervisionproblemthatemphasizesinferenceasmuchasmeasurement.
Bayesian reconstruction of 3d human motion from single-camera video Three-dimensional motion capture for human subjects is underde- termined when the input ...
Bayesian reconstruction ... 暂无评价 8页 免费 ...from video sequences 2.1 Camera orientation ...estimation of 3D human motion from an image ...
Bayesian Reconstruction of 3D Shapes and Scenes From A Single Image It’s common experience for human vision to perceive full 3D shape and scene from ...
Bayesian reconstruction ... 暂无评价 7页 免费 Human Motion Reconstruct...... This report gives an overview of 3D reconstruction from video sequences. ...
Bayesian reconstruction ... 暂无评价 7页 免费 ...Capture from Single or Multi-Camera Video Sequence...estimation of 3D human motion from an image ...
Numericals for total variation-based reconstruction of motion blurred images_专业资料。Aplp 0. Cis 知 neeE 21 2()37330053:6-7 Numeiasfrttlvrai...
Applying Prior Knowledge to Reconstruct 3D Human MotionApplying Prior Knowledge... Bayesian Reconstruction of 3D Human Motion from Single-Camera Video Nicholas...
Human Motion Reconstruct... 暂无评价 8页 免费 Bayesian reconstruction ......introduced a more accurate hand tracking system using a single video camera....
Bayesian reconstructio... 暂无评价 8页 免费 ... 3D Modeling for Capturing Human Motion from ...We propose a method for 3D reconstruction of ...
热门文档