The Research Of Human Face Detection And Tracking Algorithm In Video

For video sequence large volume, complex background and serious face block to achieve the demand of algorithm with good read-time, low error detection and great robustness, we carry out in-depth study based on face detection, face tracking and color histogram matching and eventually form a complete algorithm of human face detection and tracking from video. Our work as follows:The research of skin detection and AdaBoost detection algorithm based on Haar-like feature. In color space model (YCbCr) , using skin color detection and segment to remove the background area, combined with mathematical morphology to pre-process the skin color binary image, which great improved face detection rate of AdaBoost algorithm.Using the results of the improved human face detection algorithm to initial human face region of interest to improve CamShift tracking algorithm into automated tracking algorithm. And we also use mathematical morphology operations to process target human face back projection image, which can remove the small bright area of the back projection image (salt noise), enhanced system stability and improve the efficiency of extraction of human face.As to the error tracking face results from the serious block face in the video sequence frame image, using the histogram matching method to find a serious error tracking of video frames, comparing histogram matching methods by our experiment, compared to its advantage and disadvantage to select histogram intersect method to improve the real-time feature and robustness of face tracking.