Research On Face Detection Algorithm

Face detection and recognition technology is an important method for identity authentication using biometrics. To construct a usable and practical system, this paper researches the face detection and tracking problems under complex background and varying illumination after reviewing the history of face detection.An algorithm for detection human face in color image is described in this paper. First, a linear threshold rule is constructed according to the face color model under YCbCr space. With an ellipse model in CbCr space the color image is translated into binary image by differentiating skin pixels from non-skin pixels. Then morphological algorithm is used to take out the discrete points and unite the over-intersected region. Face area is gained by horizontal and vertical projection in binary image.Because the camera is controlled in real time according to the target information, the background and the zoom face in are changeable always. This paper uses motion estimation to track face. For the first frame, face size and location are obtained after face detection in full scene. In the following frames, face detection is processed only in the estimated region. If the target is not found, full scene detection is adopted. So the detection rate is higher and the time needed is lower to reach the real-time demand easily. At the same time, the control information is sent to camera control unit to adjust the zoom and the direct of the camera according to the target data. The application also supervises the target data in real time to stop the camera motion in time.