Face Detection And Recognition Methods Research Based On Fast-Adaboost Algorithm

With the rapid development of society, identity authentication has become an indispensable part of people’s life, so people put forward higher requirements on accuracy, safety, reliability of identification. Because of the specific physiological characteristics of human (fingerprint, face, iris, etc) have individual difference, which provide the possibility of identity authentication, and make people pay more attention to biological feature recognition. The difference of human face is the most intuitive description between the person, and face acquisition has non-invasive, so face recognition has become a very hot research topic in the biometric recognition field.In view of the importance of face recognition in real life, this paper will systematically study on face recognition. The main content of face recognition includes:face detection and face recognition. As a basic part of face recognition, face detection is vital to face recognition. However, in terms of the current research status of face detection, the algorithm still has many defects, such as degradation. So the intensive study of face detection problem is very necessary. As an application direction of face detection, face recognition is very widely used only when it is very practical. However, many factors affect the recognition rate, and the robustness of the existing system is very poor, the accuracy of face recognition remains to be further improved.After carefully reading a lot of literature about face detection and recognition and finishing relevant experiments, this paper summarized the main problems in current face detection and recognition, and focused on solving these problems. The research contents and innovations as follows:(1) This paper analyzed the current research status of face detection and recognition, then determined the research direction and key issues to be resolved through the comparative study of the existing algorithms.(2) This paper introduced the fast-AdaBoost algorithm, the concept of integral image, and rectangular eigenvalue. It improved and perfected the types of rectangle features aiming at the deficiency of existing Haar features, focused on the principles of fast-AdaBoost algorithm for training weak classifier, gave the difference of training weak classifier between fast-AdaBoost and AdaBoost algorithm, designed flow chart of the fast-AdaBoost algorithm, and programmed the algorithm on the MATLAB platform.(3) This paper summarized the problem of the fast-AdaBoost algorithm, introduced LAC theory to solve the degradation of sample weights, put forward face detection algorithm which based on LAC dynamic sample weight update, designed and trained a cascade face detector, and gave a concrete search method for human face in images. The experiment result shown that the improved face detection algorithm is reasonable in this paper.(4) Finally, this paper studied the principle of face recognition based on PC A algorithm, conducted experiments on the ORL face database, designed flow chart that satisfy both the static face recognition and the real-time face recognition system, analyzed the experimental results, summarized the shortcomings, and forecast the future development trend of face detection and recognition.