Research On Face Recognition Algorithm Based On 2DPCA And PCA Feature Extraction

We define Face recognition technology that face images are analyzed to obtain the effective information to identify the identity by the computer, which involves disciplines such as pattern recognition, machine learning, computer vision and so on. And it has been widely used in the entrance guard system, criminal identification, video monitoring and other fields. But face recognition is often affected by the change of light and angle. At the same time, in the case of a higher recognition rate, we can reduce the recognition time, which will restrict the development of face recognition.In Face recognition process, the first stage is to do image preprocessing, the second one is the feature extraction stage, the last one is to identify stage. In this paper, the research emphasis lies in the feature extraction. Based on classical PCA method,2DPCA and modular2DPCA are studied, and later, put forward the fusion modular2DPCA and PCA face recognition method,and an independent feature extraction method based on modular2DPCA. The specific research work are arranged as following:First in the feature extraction stage, respectively we introduce the PCA,2DPCA, right compressionR2DPCA, left compression2DPCA and modular2DPCA. For PCA, as the dimension of the original image is very high, less sample data, which often can appear SVD problem, the problem of small sample. And in the whole process of feature extraction and recognition, it will time-consuming. In order to solve this problem, someone has put forward a two-dimensional projection face recognition that is2DPCA. The first step to PCA algorithm is that make an image into a vector. But the amount of calculation is very complex while calculating the total scattering matrix. On this basis,2DPCA processing method is proposed. Before extracting image feature,we needn’t divide images into image vector. This method can quickly reduce the dimensions of the identification features. Again due to the distribution of human face features, obvious local characteristics, therefore, someone has proposed face recognition method based on modular2DPCA.Secondly we introduce the modular2DPCA and PCA fusion method,and an independent feature extraction method based on modular2DPCA. Because the modular2DPCA obtains that the number of feature is still large, and it still have a certain correlation in feature extraction parameter, which influences on the speed of classification. In order to solve this problem, this paper proposes modular2DPCA and PCA fusion method. Again, considering the characteristics of each sub image therefore, we carry on the independent feature extraction, known in this article,an independent feature extraction based on modular2DPCA.Finally, the experiment in ORL and Yale face library used for validation, experiments show that the proposed combination of face recognition algorithm has great recognition rate; At the same time, while illumination and expression change obviously, an independent feature extraction based on modular2DPCA has a better robustness and higher recognition rate to achieve the desired effect.