Research On Related Problems Of Face Recognition System

Automatic face recognition is an important biometric identification technology, whichwill be applied extensively. Compared with other identification methods, face recognition hasthe advantages of interactive friendly, high reliability and good concealment etc. Therefore,research of the problem of Automatic face recognition has not only considerable significationof application but also very important theory meanings. Although the current face recognitionsystems can achieve better results, it’s still affected by expression, illumination, pose andother factors. For this reason, The Automatic face recognition is studied, the following worksare completed.(1) To solve the problem of traditional AdaBoost algorithm has a high false positive rateproblem and improve the real-time of it. In this paper the fast face detection algorithm basedon skin color and AdaBoost is proposed. First skin model is used for segmenting skin area andnone skin area to achieve a pre-estimated face position, then the segmented skin area usingAdaBoost algorithm to further confirm face. The method in this paper greatly reduces the useof sliding window and shortens the detection time.(2) To enhance the ability of Local Phase Quantization (LPQ) operator insuper-resolution of face images, Local Shearlet Phase Quantization (LSPQ) is proposed,which introduces the method of Shearlet transform into LPQ. In this paper, the histogramsequence of Local Shearlet Phase Quantization (HLSPQ) is extracted from the magnitudes ofShearlet coefficients, which is used for face description. The HLSPQ operator has a betterperformance in extracting the texture features under illumination or noise effect.(3) To solve the problem of the Pyramid Histogram of Oriented Gradients has a poorperformance in describing the shape of face with noise or abrupt intensity changes, theimproved PHOG feature used in the clear outline face recognition to describe the furtherrefinement of the local structure of the face. The improved PHOG feature also can goodrestrain to noise or abrupt intensity changes(4) Constructing the framework of fully functioning face recognition system based onnetwork camera, this article is to offer the design project and realization ways of this system.At the same time in the process of the realization of the entire system, using Access to build aface database which can automatically add, modify and delete training samples, providing abasic for further investigation of face detection and recognition algorithm.