Design of face tracking system based on rotatable camera

Abstract

In order to solve the problems of detecting and tracking human face in multiple-view and under complicated background by using camera in real-time, this paper puts forward a new way that combines the multi-view face detection algorithm and face tracking algorithm to control a rotatable camera. The built-in system is able to track a human face through a camera, in which AdaBoost algorithm and Continuously Adaptive Mean Shift (CamShift) algorithm are adopted to detect and track the face, respectively. For reducing the face detection and tracking time, the preprocessing methods of skin region detection and image sampling are used. The latter one showed improved performance in face detection and tracking far better than the former in real-time. The experimental results show that the system can keep the tracked human face in the middle of visual field of the camera approximately with high accuracy and low cost.