Research And Implementation In Face Recognition Based On Multiple Classifiers

Automatic face detection and recognition is one of the most attention branches of biometrics and it is also the one of the most active and challenging tasks for image processing, pattern recognition and computer vision. It is widely applied in commercial and law area, such as mug shots retrieval, real-time video surveillance in security system and cryptography in bank and so on. It is closely linked with the field of Man-machine interaction Perception.Firstly, this paper analyzes the technology of face recognition. The background of face recognition, its research content, and its main methods are emphasized. This paper experiments in the ORL and YALE face database. Principal Component Analysis (PCA),Two-dimensional Principal Component Analysis (2D PCA) on facial image feature extraction are used in this paper. The result of facial image feature extraction is used for the face recognition of the following classifiers.Secondly, the in-depth study of the existing facial classification methods has been done in this paper. Then the recognition of facial image is realized by Euclidean distance classifier and Correlation coefficient classifier and BP neural network classifier and One-Against-All decomposition for SVM classifier and One-Against-One decomposition for SVM classifier. Higher recognition rate has been achieved by the methods used in this paper. In the ORL face database, the highest recognition rates are 93.5% in One-Against-All decomposition for SVM classifier,90% in One-Against-One decomposition for SVM classifier,93% in BP neural network classifier,89.5% in Euclidean distance classifier and 89.5% in Correlation coefficient classifier.Finally, in the ORL face database, by majority vote method, we combine three kinds of classifiers which are One-Against-All decomposition for SVM classifier, BP neural network classifier, Correlation coefficient classifier.The combination of multiple classifiers can to some extent improve the performance of classification and achieve better recognition effect. The recognition rate of combination is higher than any single classifier at every principal component. The highest recognition rate has achieved 94.5%.