Based On Sub-pattern Locality Preserving Projection Approaches Research For Race Recognition

With the development of computer science technology, some directions, such as digital image processing and pattern recognition, are attracting more and more researchers'attention. During the last decade, for the rapid spread of the network and informationization, identity identification becomes a more serious problem, which makes biometric technology become the most popular research in the world. Among most of the current identification algorithms, researchers pay more attention on face identification for its excellent stability, easy using and convenience. Therefore, face identification becomes the major research direction in the field of biometric identification.The background and present trends of face identification are first introduced briefly. Then I describe the basic knowledge of face identification, including the thorough research technologies, flowchart and so on. Finally, the original locality preserving projection method combined with spatially smooth subspace learning method is presented. Based on the presented method, I join the sub-pattern idea, whose purpose is to improve the recognition performance.The platform of this thesis is MATLAB. I do extensive experiments for face identification with different technologies, based on which, detailed analysis is given. The delicately experimental data prove that the algorithm adopted can improve the performance to some extent, which is a cornerstone for future biometrics.