Face Detection And Tracking Based On Skin Color

Research efforts in face processing include face detection, face tracking, face recognition and expression analysis. The face detection and tracking is key technology of the field of face information processing. It's widely applied in the fields of automated face recognition system, based on content index, video monitoring and human-machine alternative technology. At the same time, as the first step of human face recognition and other face applications, human face detection is very important .so human face detection has become a critical issue in the research fields of computer vision, pattern recognition.In this paper several research works are given as follows:First the four general ideas of face recognition are pondered upon and analyzed, this thesis brings forward a reasonable and effective methods on complexion pick-up. It proves that this methods reduces much time .In the end, the thesis puts forward a novel average face template based on edge detection which is used to match with the binary image reserving the details of human faces. It greatly reduces the influence of complex scenes.Second the methods of image segmentation are summarized and analyzed. An algorithm of image segmentation is developed based on self-adaptive threshold value. With the difference algorithm of image sequence, I can eliminate the background and extrude the moving object in images. Then hair clustering analysis, face geometry character applied to verify very candidate region to get real face regions. The detected face is used as matching template in face tracking.Finally face real-time capture, face detection and tracking have been realized by video camera. The video camera is connected to computer by USB. In this paper, all algorithms were accomplished by VC++ in Windows XP environment. Experiments results shows that this system is reliable .The detect-rate of algorithm reaches 90% through detecting human faces on 40 test images. Except consuming more time, this novel methods is prior to the common methods based on complexion and template matching.