Recognition, Based On Classified Information Nmf Image Reconstruction

As one of the research contents in biometrics recognition technology, face recognition is the new frontier in the fields of pattern recognition, artificial intelligence and computer vision. Compared with the other methods of recognition using fingerprints, hand shape, retina, signature, voice, iris recognition or other features, face recognition technology is universal, non-invasive, accessible, accept easy, stable. Face recognition has been widely used in criminal identification, banking and customs control system, information security and other fields.Face recognition technology uses computer processing human images, analyzing effective identification information contained in the images, and then removing the invalid redundancy, noise and other information, in the end compared with face images in the database, in order to identify the classification of the face image. Currently there are two main research directions in the face recognition:1, based on the overall research methodology, the main consideration in this direction is the whole property contained in the face image, through extracting the global features to recognize faces. Including the Eigenfaces method, neural network methods; 2, feature-based analysis, in this direction, extracting the geometry, algebra features to recognize faces. For example:Based on the extraction of the eyes, ears, nose and other parts of the formation characteristic and then use this feature to get eigenvector.This paper considers the classification of face images information, on this basis, reconstructed face image, using non-negative matrix to extract the features of face images. Compared with PCA, BPCA in ORL face database, we improve the recognition rate and recognition speed. The experimental results show the effectiveness of the method.The mainly research contents of this paper are as follows:1, This paper describes the development of face recognition, the process of face recognition, the current major challenges in the face recognition, and then describes the status of face recognition briefly. I makes a detailed analysis of the field which has been widely used in face recognition techniques, such as PCA, FDA, SVD, NMF and other face recognition methods.2, Based on the symmetry and block characteristics, I uses PCA to recognize classification of the face image. Firstly, I makes a detailed analysis of PCA method and uses this technology to reduce dimension, Secondly, blocks the face image to improve the samples number. On this basis, lastly, I considers the symmetry features of the human face image. The experiment compared with PCA method and BPCA method proved the effectiveness of this method.3, Through study the non-negative matrix factorization, I uses this algorithm to get the coefficient matrix and basis matrix and then, considers planar to get more image planes of the face, at the same time combined with NMF methods to identify the face category.Lastly, the experiment compared with SVD+NMF proves the effectiveness of the proposed method.