| Face recognition is a research focus in pattern recognition, image processing and other subjects. It is widely used for authentication, investigation, video surveillance, intelligent robots, medicine and other areas. Face recognition has wide application and business value. Facial feature as a biological characteristic, compared with others is direct, friendly and convenient. Facial features used for authentication is easier for users to accept.Face recognition has made a great development in the past several decades. There are many scholars at domestic and abroad studied on face recognition. However, facial features, such as non-rigid, facial expression changes and other features, make the practical application of face recognition confront enormous difficulties. In the paper, based on the analysis and research on academic and reports all over the world on face recognition in recent years, combined the characteristics of rough sets and artificial neural network, and proposed a face recognition method based on rough set and neural network. And, in the process of implementation and certification of the method, this paper proposed a new optimized method of PCA face recognition simultaneously.New optimized method of PCA face recognition is proposed for variable illumination conditions. In pre-processing stage, the new method applies the linear transformation to increase each gray image's contrast grade and brightness, then, handled the pretreated images according to the classical PCA face recognition arithmetic. In the recognition stage, the first three principal components which reflected the illumination conditions were given a suitable weight, in order to reduce the proportion they accounted for. The experimental results show that the new method is effective in reducing the illumination.Face recognition based on rough set and neural network was proposed for the shortcoming of high dimension of PCA face recognition and low recognition rate for non-training samples. On the basis of the former pretreatment, using the rough sets, it reduced the facial features which handled by PCA method, and extracted the strong features of classification. Using the rough sets, identification accuracy can be achieved in the same circumstances, and also can effectively remove the redundant information. Then, after finishing attribute reduction, the extracted facial features were input to the neural network to train the network and recognition. Using the nonlinear mapping and parallel processing characteristics of the neural network, the method enhanced the neural network's generalization on face recognition. Experimental results show that the recognition rate got some increase. |
