The Study Of Improved PCA And BP Neural Network Face Recognitionh

Face recognition technology is a hot research topic in pattern recognition, imageprocessing and other subjects. After decades of development, face recognition technologyhas made many achievements, but it still has some distance away from full practical use.Based on this, the study of practical face recognition is an important direction ofdevelopment for face recognition technology. In this paper, by using the fusion ofmulti-data and different methods, I try to study a more practical way of face recognitionthat is applied and can also effectively improve the face recognition rate based on thecurrent face recognition algorithm. The main research work in this paper are as follows：Firstly, I study the face recognition system combined with classical PCA and BPneural network. Using the ORL face database as sample, though the design of algorithmsteps, I carry out the simulation experiments. And the simulation results indicate that thisface recognition system has a low recognition rate as a shortcoming. That is because PCAcan not extract facial features effectively alone. So must process the image before the stepof PCA to highlight the main features of information and hide secondary information.Again, for the shortcoming of the face recognition system combined with PCA andBP neural network, I study the improved Gabor program. The simulation results obtainedby the improved Gabor program can effectively improve the recognition rate. But it alsoincreases the dimension of the final feature information for identifying. So the furtherfeature information extrication is needed.Finally, I study the improved program of rough sets for the shortcoming of Gaborimproved system. The simulation results show that the improved systems by rough set canreduce the dimension of the feature information effectively and not affect the recognitionrate. My study of the way by combining of Gabor wavelet, PCA dimensionality reductiontechnology, rough sets reduction and BP neural network technology can not onlyeffectively dig out the local characteristics of the face image, but also has a goodrobustness in the face image with illumination, different expressions and postures. Thehigher recognition rate of the proposed face recognition method tested by ORL proves that this method is feasible and has practical value.