Research On A Real-time Method Of Face Detection For Smart Video Surveillance

In the recent years, the Public and Personal Security is becoming a prominent issue of modern society. With the increase of terrorist threats and high level of criminality, video surveillance, as a popular security tool which elevates the public security level in some extend, is becoming ubiquitous.Traditional video surveillance system only has limited effect on crime, for the recorded video data generally is used after the paroxysmal events. Oppositely, the smart intelligent video surveillance system attempts to detect, recognize and track certain moving objects from sequences, to understand object behaviors, and predict the activities of the objects. Under some situation, human face is of great importance in surveillance system.In this paper, a novel method of face detection for video surveillance based on the AdaBoost approach is proposed. Compare with traditional face detection methods, our method makes use of a background model and spatial-temporal information in video and achieves better real-time performance. Furthermore, we achieve lower false detection rate by a fast and universal skin colour model.