Face Recognition Hybrid Algorithm Based On Modified PCA

As each person's face has its own peculiarities, it is simple and convenient to operate face recognition system, and this kind of system is without invasive, making face recognition technology become an important research direction of biometric identification, and attract more and more research groups'and research institutes' attention on it. Face recognition technology involves digital image processing, pattern recognition, computer vision, neural networks and other research areas, has important research values, The technology can be widely applied in many areas of the real world(such as security, digital surveillance, document authentication and access control, etc.), and therefore it has great application value.This paper is started by face recognition based on principal component analysis thought, first deeply analyzes the basic principle of PCA algorithm, designs improved PCA algorithm and improved PCA+LDA algorithm based on PCA algorithm. Improved PCA algorithm combines PCA algorithm with the processing of enhancing local mean and standard deviation of image algorithm to improve the recognition of human faces when the light and robust expression changing. When PCA algorithm used for face recognition to extract facial features, the extraction is the most expressing features, rather than the most discriminative features, a face recognition method based on improved PCA+ LDA is designed in this paper. Second, deeply study the basic principle, advantages and disadvantages of 2DPCA algorithm, and designed a face recognition method based on improved PCA+2DPCA, first using method of 2DPCA to extract feature of the sample images, getting the corresponding feature matrix of images, and then using PCA to reduce its dimensions, which can not only reducing the feature dimension of the sample image, but also improving the recognition performance of the system.With the ORL and Yale face database, the relevant experiments and validation experiments is done for face recognition methods proposed by this paper, the result shows that proposed methods have good recognition performance, have certain robustness on the uneven illumination and expression changing human faces.