Real-time head tracking system with an active camera

Human head detection played an important role in applications such as video surveillance, perceptual user interface, face recognition and people tracking. In this system, an improved moving object detection algorithm was used to start the head detection and tracking process, which was based on the gradient and the color information. Firstly, we decreased the image size through a sampling method, then we detected skin regions over the entire image, and generated a searching space from them, after that an ellipse model was used to detect the head in the searching space, and the detection result was reflected to the original image, finally the final result above was being used to control the active camera. The evaluation of this system´s performance was shown in the last part of the article. Experiment result shows that this system runs in real-time on a standard PC, being robust to changing background, partial occlusion, clutter, face scale variations, rotations in depth, and fast changes in subject or camera position.