Research Of Color Face Detection Based On 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 persons' faces. To develop a robust automatic face recognition system is a target in the field of Machine Vision. And the first step of recognition is the detection of the face, which works as the foundation of the availability and efficiency of the whole system. Recent years, owing to the greatly potential requirement in the applications of security surveillance, image search based on contents 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 literatures, surveys and research papers concerning up-to-date techniques of face detection and face recognition are read and analyzed. Some hot issues about face detection are discussed. The research of the first important step of building an automatic face recognition system—face detection are attempted. Experiments indicate that the methods of face detection proposed 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 aspects:1. Traditional edge detection algorithms are introduced in human face detection applications. The comparing of several common algorithms' advantages and disadvantages is made through the results of them. Two potential algorithms—Canny and Sobel—are selected and a hybrid method combining these two are attempted to adopt to the application of face detection. Comparative ideal edge lines of faces are detected, which have meaningful indication to next step of detection.2. Face detection algorithm based on the model of faces' skin color is carried out by using the color information of color image. YCrCb color space in which chroma and luminance are separated is adopted because of its high suitability to the face detection application comparing with other color spaces. The results show that skin color information plays a practical role in the fast face detection.3. The system of face detection based on SVM is developed. SVM theory andalgorithm are deeply studied. Through the process of the achieving of fast multi-scale face detection, the thesis mainly focuses on how to overcome the time-consuming feature of SVM by using the skin color information of images to reduce the anticipative areas, heuristic search method to increase dealing speed and PCA translation to decrease the dimensions of face samples. Experimental results show that the algorithm based on machine learning theory has better theory value and practical value.4. Based on these experimental results, the disadvantages and merits of small computing features, like skin color, edge and so on, are analyzed. The practical perspective of theory base on Statistics like SVM is presented.Face detection, including recognition is a highly practical application. The particular demand of application must be taken enough attention on the process of detection and recognition of human face, which is such a complicated high-dimension pattern. Integrated methods are frequently employed to achieve the accuracy, efficiency and real-time ability of the system.