Based On Adaptive Color Model And The Geometric Characteristics Of The Face Detection

Face detection is a key technique processing and analysis in human face information, and it mainly come from face recognition, and it is an important element in automatic human face recognition systems. It has become a very active and popular research topic in computer vision and pattern recognition. It plays a prominent role in video detection and tracking content-based retrieval facial expression recognition video encoding entry-exit safety inspection visual monitoring, and has the actual applied value.Skin-color feature is a very important information in color images,it does not depend on the details of facial features and face rotation angle scale of the face, and has relative stability. Skin color distribution showed good clustering in color space, skin color has quick processing handling speed, skin color segmentation higher efficiency, this make skin color segmentation become a kind of very important method for face detection. Using skin color feature information can quickly remove lots of background and derive the target skin area.The primary research works of this paper are as following:1,Face Detection were classified and summarized at home and abroad , discussed the face detection technology in several common color space, advantages and disadvantages of various color spaces and whether there is the existence of the optimal color space, mainly introduces several methods of established skin color model and their characters.2,According to the color space and color model analysis, we propose a reliable adaptive skin color modeling approach, It is based on the fact that color distribution of a single-colored object is not invariant with respect to brightness variations even in the Cb-Cr plane,and does not ignore the influence on brightness Y component in YCbCr color space. Firstly, According to statistics, we take the luminance Y by ascending order and divide the total range of Y into finite number of intervals, N. Collect pixels whose luminance belongs to the same intensity interval and calculates the standard deviation and the mean value of Cb and Cr with respect to Y, then we use the above data to train the BP neural network, then we get a adaptive skin color model.3,Based on the improved algorithm of skin color segmentation,candidate face area can be determined by face feature and the geometry relationship among the salient components of the face,And then use Mosaic to confirms the candidate face area.4,When locating the eyes'centre, First use the horizontal integration projection method to determine the upper and lower facial area boundary, second, use the vertical integration projection to determine the left and right boundary of face and propose a method of location the eyes center based on the circle and ellipse then according to the human eye centers to normalize the face, the normalized face include the main features of the faces.5,The results of simulation show that the proposed new algorithm has fairly good effect; The result of experiment has validated the correctness of the design and research in this paper and provided significant guidance to the practical application.