Study On The Methods For Fast Human Face Detection And Tracking

Detecting and tracking human face robustly and fast is the most important issue during the study of the face detection and tracking. To resolve the key issue, some new methods for fast locating and tracking human faces were proposed in the thesis and listed as follow:1. Fast approach to the segmentation of human faces based on particle swarm optimization algorithmThe thesis proposes a novel and fast approach to the segmentation of human faces in color image under complex background. In order to find a face fast, unlike the traditional method in which all the points are searched, the proposed approach utilizes particle swarm optimization algorithm to mark and cluster the skin-like pixels. Experimental results show that the proposed approach is fast and has a high detection rate.2. Adaptive face tracking based on mean shift algorithmThe thesis proposes an adaptive face tracking method which an adaptive facial orientation template is adopted based on mean shift algorithm. It can describe the face pose more accurately than the traditional mean shift tracking method when the facial orientation changed. Experimental results show that the method is more efficient to adapt the facial orientation.3. Multiple faces tracking based on mean shift algorithm combined with Kalman filterThe thesis proposes an approach for multiple faces tracking based on mean shift algorithm combined with Kalman filter. For each tracker, the approach takes the mean shift vector as the observation. Experimental results show that the tracking speed is about 20 frames per second (fps) for 320×240 video size.