Skip to Main Content
In this paper, we present a novel system that acquires clear human face or head in video stream from a single surveillance camera and applies several state-of-the-art computer vision algorithms to generate real-time human head detection and tracking results. The detection results are obtained using histogram of gradients (HoG) feature powered by GPU and the false positives are filtered out by motion and appearance likelihoods. Then all the true positives are tracked by a track algorithm based on earth mover's distance (EMD) and SURF points. Our experiment results show that the proposed system makes it possible to accomplish robust detection and tracking performance in most surveillance scenes.
Date of Conference: 29-31 May 2012