Markerless Motion Capture with Multi-view Structured Light.

摘要: We present a multi-view structured light system for markerless motion capture of human subjects. In contrast to existing approaches that use multiple camera streams, we reconstruct the scene by combining six partial 3D scans generated from three stations surrounding subject and operating in round robin fashion. avoid interference between projectors through time multiplexing synchronization across all cameras projectors. Assuming known multi-camera projector calibration parameters, generate point clouds each station, convert them surfaces, merge into single coordinate frame. develop algorithms dynamic geometry using template itself. Specifically, deform frame captured iteratively aligning bone skeleton. This is done searching correspondences source geometry, solving rotations bones, enforcing constraints on rotation prevent taking anatomically unnatural poses. Once sequence processed, textured color images systems. show effectiveness our 50-second moving subject.