Tracking End-Effectors for Marker-Less 3D Human Motion Estimation in Multi-view Image Sequences

摘要

We propose to track the end-effectors of human body, and use them as kinematic constraints for reliable marker-less 3D human motion tracking. In the presented approach, we track the end-effectors using particle filtering. The tracked results are then combined with image features for 3D full pose tracking. Experimental results verified that the inclusion of end-effectors' constraints improves the tracking performances.