A Study On Face Detection And Face Tracking Techniques For Intelligent Video Surveillance

Face detection and face tracking have a wide range of applications; they were usedin video surveillance, biometrics, human-computer interaction, image retrieval, videocoding, etc. This paper studys face detection and face tracking for intelligent videosurveillance application. Firstly, the development of the face detection and facetracking technology is summarized and analyzed, and then the improved AdaBoostface detection and CAMShift face tracking algorithms are developed. The details areas follows.(1) A real-time face detector based on Gentle AdaBoost algorithm and NestingCascade Structure has been proposed, and experimental results have shown that thistraining algorithm can significantly reduce the number of weak classifiers in thecascade classifier, which dramatically improves the face detection speed and slightlyimproves the detection accuracy. Moreover, the OpenMP parallel programmingtechnique is exploited to speed up the training process.(2) Based on the aforementioned face detector, a fast face detection system withmotion detection pre-processor and complexion verification post-processor has beenproposed. In which motion detector is used to shrink the searching area, and skinverification is used to filter the texture interferences. Besides, the complexionverification processor is adaptively switched on and off, according to the averagebrightness of face candidate area, to further reduces the mistake detection rate.(3) In order to improve the robustness of the CAMShift tracking algorithm, thispaper introduced a Haar feature pre-fiter to eliminate the similar skin color area.Experimental results have shown that this algorithm can effectively improve therobustness of the face tracking.