Model based human pose estimation in MultiCamera using weighted particle filters

Abstract

Human motion capture and human pose estimation is one of the most active research area in computer vision. Pose estimation refers to the process of estimating the configuration of the kinematic or skeletal articulation structure of a person. Recent research area refers to multi-camera 3D pose estimation that is very useful and more robust to self-occluding and ambiguous. These approaches use fusion of features that extracted from each camera in different views. In this paper we proposed a novel fusion algorithm based on multiple particle filters in decision level fusion. We use particle filter to tracking the body parts on each camera. Then we use decision fusion by weighting estimation results from each camera. Experimental results in HumanEva dataset and comparison of our work to one of the recent works in this area show the better accuracy especially in some difficult situations. © 2012 Springer-Verlag.