Research On Face Detection Method Based On The Eye Positioning And Support Vector Machine

The automatic recognition and detection of human face is one of the most interesting and challenging topics in the fields of Artificial Intelligence and Computer Vision. It is the aim of many researchers and scientists working in this field to make computer own human's natural ability to remember and recognize person's face. Research robust automatic face recognition system is a substantial target in the fields of Machine Vision, and the first step of recognition is the detection of the face, which as the foundation of the availability and the efficiency of the whole system. In recent years, owing to the greatly potential requirement in the applications of security surveillance, and image search based on content and so on, face detection is developing as a systemic and independent research branch, attracting more and more interests of researchers.A great amount of essays, surveys and research papers concerning up-to-date techniques of face detection and face recognition in recent years are read and analyzed by this paper. Some hot issues about face detection are discussed. The research of the first important step of building an automatic face recognition system——face detection is done. Experiments indicate that the methods of face detection considered in this paper are reasonable, showing a certain degree of theoretical and practical value. The research work of this paper mainly includes the following several respects:Firstly, on the basis of description the theories and the technique of the face detection system, this paper analyze the arithmetic of the classic of the face detection, and make the contrast of them. And summarize their advantages and disadvantages.Secondly, according to the complexion feature independent of change of the face detail, the rotation and the expression and so on, the complexion feature is easy to distinguish from most of the background colors, and has the good stability. This paper tries an eye positioning method based on the face geometrical feature and its skin gray. The eye positioning method has the effective balance to this contradiction between the detection speed and the stability. And it also has an adaptive to the background and the size of the test image. Subsequently, the paper uses a "structure classifier" that combination the skin color model to optimize the region of the eye positioning.Thirdly, based on further analysis of the theory and arithmetic of SVM, this paper provides a new method of face detection. The method divides one face image into three parts and makes use of the "learning again" way of the system. The analysis show that this method is effective and remarkably.Finally, based on the above related theories and method, this paper combination the eye positioning, the optimized method of structure classifier and the SVM method, and makes the face confirmation. The putout results show that the new method is satisfying in accuracy of the face detection and the face positioning.