The Face Recognition Technology Based On Principal Component Analysis

Principal Component Analysis (PCA) is a widely used face recognition method in recent years. The PCA method requires changing the image from the two-dimensional matrix into a one-dimensional vector, then constructs the covariance matrix which has a huge dimension, and calculates the eigenvalues and eigenvectors, which is very complex. In recent years, the Modular Principal Component Analysis (Modular PCA) and Two-dimensional Principal Component Analysis (2DPCA) face recognition methods have been improved to varying degrees in terms of recognition rate and recognition time respectively.In this paper, we studied the Modular PCA and 2DPCA face recognition methods based on PCA face recognition method, and presented an improved 2DPCA method which is called Modular Weighted 2DPCA, the main research work in this paper is described as follows:At first, we introduce the principle and implementation process of the PCA face recognition method. Because the traditional PCA method directly on the global information of the image, and extract the global features, while ignore many local details which are the necessary information for pattern recognition. When people's face has a great change in expression or lighting conditions, the recognition results are unsatisfactory. Therefore, we studied a face recognition method called Modular PCA method. Fristly, we divide every facial image into equal size sub-images, stretch all the sub-images into a single sub-vector, then extract sub-vector feature further.Then we introduced the basic principles and implementation process of the 2DPCA face recognition method. Because the method of 2DPCA discards all correlation information among the columns of images, thus it is difficult to characterize the local features of human face. To overcome the shortcomings of 2DPCA, this paper presents a Modular Weighted 2DPCA method.Finally, we verify all the methods that proposed above by experiments in the ORL and Yale face database, and make a simple system for face recognition. The results of all the experiments shows that when a greater expression or illumination changes, the Modular Weighted 2DPCA method had better robustness, and achieved a higher recognition rates.