Research On Face Detection And Tracking

Face detection and tracking is a foundational and key technology in the facial analysis,thus can be applied in many fields, such as video surveillance, human-computer interaction,biometrics, content-based image retrieval, image coding, etc.Regarding the face detection, this paper mainly focuses on color-based face detectionalgorithm and AdaBoost face detection algorithm. Inspired by the integral image, the shincolor binaryzation integral image is used to calculate the skin pixel ratio in detecting windowrapidly, and skin information is introduced into AdaBoost face detection algorithm to reducethe rate of error detection in color images.Regarding the face tracking, the Mean Shift tracking algorithm and the CamShiftalgorithm are investigated in detail. The CamShift algorithm is an improved algorithm basedon Mean-Shift. Since the algorithm can robustly track target of different shape and size withthe immunity against illuminant fluctuation and noise interference, and has low CPU load, itcan be served as an efficient human-computer interface.However, CamShift performsunsatisfactorily when flesh-like interference and occlusion occur. The update of trackingobject and accessory information are used to enhance the robustness of CamShift, and facedetection based on AdaBoost is used to initiate the searching window automatically. MultipleCamShift trackers are used to realize multiple faces tracking.Finally on the application environment based on VC++6.0, the experimental system ofautomatic multi-faces detection and tracking is developed.Experimental result has testifiedthat the improved algorithm has stronger robustness and efficiency.