Face detection is the use of computer to find and locate the face, get the size, location, number and other information of the face in image, further can extract face characteristics. In recent years, It is considered to be one of the important research subject in pattern recognition and computer vision areas. This paper based on the study on the basis of the previous achievements, face detection and tracking algorithm for detecting and tracking human face in complex background is proposed based on skin color, The main research works are composed of three aspects:(1) Skin model. On the basis of the common color space YCbCr, this paper uses two YCbCr and HSV color space combined to establish skin model. In the image segmentation, we adopted mixed skin color and improved genetic algorithm to automatically select the threshold of Sobel edge detection algorithm and mathematical morphology method. Experimental results show the proposed algorithm enhance the reliability and examine a person's face position availably.(2) Location of facial characteristics locates the position of human's eyes and mouth. In the eyes segmentation, A new approach is adopted base on an improved genetic algorithm to automatically select the threshold of Sobel edge detection algorithm to detect image edge, then use mathematical morphology method to detect the position of face feature area, At last, human eyes regions are located accurately by using the method of calculating image block gray gradient density. The method is a small amount of calculations, simple and effective. Mouth region are obtained by according to the color of mouth, eyes and a prior knowledge. The human face are located accurately by according to the triangular relationship of the human eye and mouth. Experimental results show that this method can locate human face simply and accurately.(3) In the face of the tracking part,point motion prediction algorithm is proposed to estimate the approximate location of facial movement to reduce the search area. After determine the search area in probably,two orthogonal tracking model is adopted to track face, one of which is a genetic algorithm to optimize the edge gradient track model, the other is internal color pixel statistic histogram tracking model. Experiments show that this algorithm of face tracking has the very good continuity, real-time. |