Inference of 3D Human Body Posture from Multiple Cameras for Vision-Based User Interface

摘要

In this paper we presents a silhouette-based method for 3D human body reconstruction from multiple views. The silhouettes are extracted automatically from the captured frames using a background-learning approach. The multiple views are integrated using the epipolar geometry relating the pair of synchronized video streams. The 3D model of the human body is derived from the integration of the median axis of the silhouettes along with a Generalized Cylinders description of the 3D shape. The proposed approach is validated by a set of experiments capturing various human postures.