3-D Human Tracking Based on Particle Filter

A 3D human tracking method based on particle filter is introduced in this paper. In this method, firstly the face area of the human is extracted based on a skin model and template matching with standard human face from recorded 2-D video images. Then 2-D face positions in the video images are converted to 3-D world coordinates through the calibration parameters. Finally these 3-D position estimates are fed to a 3-D visual-tracker, and both position and velocity of the human can be obtained from the 3-D tracker for further applications, such as teleconference and human motion analysis. Simulations are performed to confirm the effectiveness of the proposed approach.