Performance improvement of human detecting and tracking based on stereo vision

Human detection and tracking are a vital part of mobile robot. In this paper, a method of improving performance of human detecting and tracking based on stereo vision is proposed. The proposed method is to get the human´s disparity image by stereo camera, and then we can attain the human feature by image processing. Hu moment is chosen to detect human because it has the invariant character of translation, rotation, proportion. Robust human tracking is performed with Extend Kalman Filter (EKF) as it´s flexible and easy to use in practical environments. Experimental results prove that the proposed method is reasonable.