Implementations Of Face Recognition System Based On ADSP-BF561

In the 21st century, face recognition technology has been widely developed for the advantages of quick and convenient. Face recognition involves a number of important technical disciplines. As an important research subject, face recognition has extremely important theoretical value and practical value. This thesis studies the basic technology of face recognition to realize a real-time face recognition system based on video image.The face recognition system implemented in the thesis uses an ADSP-BF561 platform as the terminal and a personal computer (PC) as the server. The face detection part of system utilizes the algorithm based on skin color. MATLAB is used to obtain the simulation results. The image is pretreated through the "reference white" illumination compensation method to compensate the color deviation. Then the Cb and Cr ellipse clustering method proposed by Anil K Jain is improved to segment skin color so as to reduce the misjudgment of color points in the high brightness and low brightness region. Further, the face region is determined via skin color based on certain rules. The algorithm is transplanted into C language. The process of the main function is explanted detailed, and is tested on BF561 platform.The face recognition part focuses on the principles and the implementation of the feature extraction algorithm based on principal component analysis (PCA), and the classification identification algorithms involving Euclidean distance and vector angle. Then the process of the main function of the algorithm is elaborates cautiously in C language. The design simulates the PCA algorithm, Euclidean distance and vector angle methods on the platform of MATLAB. Further, face images in ORL face database is trained and tested, as a result of which, the recognition rates of different classification and identification algorithms in the context of identical feature extraction algorithm and face database are obtained and compared.Finally, the overall workflow of the system is described briefly. The key points of the system task are elaborated exhaustively. Then the design is tested on the hardware platform; the test results and analysis are obtained. The design which operates normally and steady still has shortcomings for improvement. At the end of the thesis the major work and the research direction of future are summarized.