Research And Implementation Of Face Recognition System Based On Linear Discriminant Analysis

Face recognition technology is a very popular research topic in the field of pattern recognition and machine vision. It is used broadly in some institution such as customs, public security, banking and other departments owing to its good properties which are direct, convenient, friendly and easily accepted by users compared to other biometric identification technologies. It involves various technologies such as computer graphics, pattern recognition, physiology, artificial intelligence, machine vision and so on. General face recognition technology usually consists of face detection, image preprocessing, feature extraction and recognition techniques. Among these, face detection and feature extraction are critical components in a face recognition system. Therefore, this paper improved the progress in the two aspects.(1) In the face detection stage, the paper proposed a kind of comprehensive face detection method, which improved the real-time and stability in the progress of detection. In the method, first of all, we confirm the shooting environment of image according to the whole pixel information of the image. Then we locate the area of skin according to the complexion information after light compensation, and calculate the number of pixels in the skin region and adjust the image according to the number of pixels. Finally AdaBoost algorithm is adopted to detect human face. Experimental results show that this method reduces the time of face detection and gains a better performance than Adaboost algorithm in all aspects.(2) After we analyze the Principal Component Analysis(PCA) and Fisher Linear Discriminant Analysis(Fisher LDA,LDA), consider both the local features and the whole features of the image, a new LDA algorithm using DCT and LBP features fusion based on half-even images (HELDLDA) is proposed. Half-even images are used as samples in this method, then extract the LBP features and DCT features, finally we fuse the two features to do face recognition. The algorithm described above has been simulated on the platform of MATLAB and experimental results show that the method reduces the dimensions of samples, weakens the impacts of postures and facial expressions, takes both the local features and the whole features into account and it’s recognition rate is better than LDA method.(3) In order to solves the Small Sample Size problem, we introduce the2DLDA method, an improved2DLDA (HE2DLDA) algorithm based on half-even images is proposed. The method extracts facial features directly from half-even images by using2DLDA algorithm. Experimental results show that HE2DLDA method is superior to LDA method and2DLDA method in terms of feature extraction time and has the same recognition performance too.(4) An entire face recognition system is developed based on Open Source Computer Vision (OpenCVl.0) and the theory knowledge above. It has a pretty good performance through the test.