Research On Human Face Detecting Based On Prior Knowledge

With the rapid development of the social economy and the culture, the technology of fast-and-effecient authentication has become more and more important. Human face detection technology is a biometrics that can identify automativally based on a person's facial characteristics, and it has many advantages, such as direct application, friendly interface, and non-invasion, thus it is of great important research values. Human face detection is the most important step in the face recognition system, and it is also a research focus in the pattern recognition and in the computer visual field. With the wild application of the automatic monitoring and human-computer interaction, the problems of human face detection had received more attentions. At the same time, since the human face detection can be greatly influenced by background, light, gesture, expression, occlusion, noise and so on, it also became a much complicated topic.This thesis integrated all kinds of the prior knowledge about the human face detection. Firstly, it established a simple Gaussian skin color processing model; secondly, it carried out illumination compensation and dealed with the binarization separately; then it conducted rough human face detection through the geometric characteristics of human face (such as the ratio of the size of the area, the ratio of the length and width, and the features of the holes).At the end of this thesis, it introduced the integrated use of Sobel and Canny Operator on edge detecting the candidate region of human face. Here, it used Gaussian model to projection locate the candidate region of human face, and effectively detected the human informations in the pictures. In order to realize all the algorithm, this paper has developed a face detection system on the platform of MATLAB7.0. This face detection system had a friendly interface and a strong visibility, and it described the process of sofeware realization and relative models of face detection system. It also analysed and explained the experiment's results, which effectively proved that under different backgrounds, the human face detection method based on prior knowledge is more available than that based on skin color detection method.