Research On Face Recognition System Based On The 2DDPCA Algorithm

With the innovation of science and technology and the improvement of the information, people's living and learning have been convenient. But there are many very serious problems of security vulnerabilities. Traditional security technology gradually revealed defects, ranging from daily life to the whole society exist great risks. Theft of account information and data occur frequently, causing some irreparable damage. The biological characteristic is a kind of recognition method,using the human body characteristic that cannot be changed. Now getting more attention of people, the main reason is that it is difficult to be changed and to counterfeit. The most easily accepted is face recognition in biological characteristics,because it can be monitored without being found. It has been widely used in many occasions, and has important research value and application valueIn the feature extraction stage by 2DDPCA(Two Dimension Double PCA), it is a bidirectional two-dimensional feature extraction algorithm. Now the typical PCA(Principle Component Analysis) algorithm in the subspace is one of the trend, which is mainly on the basis of the linear combination(K-L transform). But the conventional PCA algorithm will interfered by the environment,no effective way to deal with the serious impact on the recognition rate. In order to suppress these disadvantages, an improved 2DPCA(Two Dimension PCA) which based on the traditional PCA face recognition algorithm has been introduced in the paper, then second feature extraction method has been used in the 2DPCA algorithm, which has been called 2DDPCA algorithm. In order to verify the performance of the improved algorithm, using three algorithms which are PCA, 2DPCA, 2DDPCA complete a large number of comparative experiments and analysis on the face database.According to the experimental data, the improved 2DDPCA algorithm which has good recognition rate and reconstruction.In the recognition phase, SVM(Support Vector Machine) is used as classifiercomplete recognition. SVM classifier has strong self-learning ability and avoid over fitting phenomenon, and it has very good classified recognition at dealing with the nonlinear classification in high-dimensional space. Combining the SVM classifier and the feature extraction algorithm, different kernel functions and parameters of the classification effect is obtained through a lot of experiments on the ORL and Yale database. Experiments show that the improved algorithm of 2DDPCA combined with SVM has better recognition performance and robustness, it will be better to adapt to complex conditions.Finally, with the help of MATLAB software can build the face recognition system based on algorithm of 2DDPCA combined with SVM. The system realizes the improved algorithm, and the recognition rate is also improved.