Study On Human Face Detection And Tracking Algorithm Based On Skin Color

Face is the most common object in the world, which exists in vedio camera phones, news, TV news ,television conference and so on. As the evoluting of man-machine interaction and the developing of biological authertication and object detection and the increasing requirements in the field of commercial securety, the technology of face detection and tracking gets studied and used more and more widely.This paper studies face detection and tracking algorithm,proposed a scheme based on the combination of color information and classifier. This scheme analyzed the distribution of face skin in RGB,YCbCr,HSV,HIS color space. Constructed a hybrid skin model based on spaces.Using this model to find skin segment, candidate area is finded.Combining with Successive Mean Quantization Transform (SMQT), SNoW classifier is trained and get the candidate region.With this detection algorithm in video sequences, proposed a tracking algrithm based on a linear motion detection of forecasting model and face verification with SVM, realized face tracking at last.In face detection stage, this paper puted forward a histogram equalization algorithm for local lighting,improved the previous compensation image. Using grayscale equalization into classic Sobel edge detection, got more smooth continuous edge curve.In face tracking,finded the center of target mass, then realized moving face tracking,and Haar wavelet transform inhanced image constrast. The result proved that the scheme using in this paper is simple, feasible and much more accurate.