Robust 3D human tracking based on Kinect

Unlike conventional human detection in RGB images or intensity images, we investigate three dimensional (3D) Human detection and tracking in depth sequence in this paper. Boundary problem is a critical issue that influences performance of detection and tracking. In this paper we design a tracking system consisting of Microsoft Kinect and stepping motor, which can guarantee people always stay in the center of camera vision. Clearly, the boundary problem can be solved efficiently. We also establish an images database, which contains depth images and RGB images for performance evaluation of the proposed auto-tracking system. Experiment results show that the proposed system is robust against variations in illumination, background, pose.