Research On The Key Technology Of The Image-based Face Recognition

Face recognition as a new biometric identification technology compared with other feature recognition technology has better acceptability; this method can be used active mode or passive way, very flexible and convenient. So face recognition technology has been applied in every field of the social reality, such as confidential departments, bank, hospital or other military key units. At the same time, face recognition technology has become one of a hot topic in the field of machine learning and pattern recognition in recent decades, and generated a large number of different patterns of face recognition algorithm, and which recognition accuracy and application range reached a considerable effect.The faces and other biological characteristics has variability and complexity, is easily affected by various external conditions, such as the age factor, attitude change, light illumination, occlusion problem; and which will bring great negative impact to the identification and authentication work. Existing face recognition algorithms can not completely solve all the problems above, and the recognition accuracy of face recognition technology can not meet the actual requirements. Therefore, how to further improve the recognition accuracy of the face recognition technology, improve and enhance the effectiveness of the algorithm for different environmental adaptability, and the operation of the algorithm is a problem to be solved.Aiming to the problems and difficulties of the current image-based face recognition technology, this paper has researched and analyzed on all aspects of the process of face recognition deeply, and found the advantages and disadvantages of the existing face recognition algorithm to deal with all kinds of problems, and on the basis of inheriting the advantages, the existing recognition methods were improved, and shortcoming of them were eliminated. Through the analysis of the theoretical and experimental results, some achievements have been obtained. Specific works as follows:The modification to the local feature extraction of face recognition, because the traditional local feature extraction methods can not fully extract the local features of the face information, this paper proposed a block complete LBP face recognition method based on the low frequency image. By using the complete LBP the image information can be extracted more completely and the block model can make some details of image were extracted and used effectively, which can well improve the recognition precision of the algorithm, At the same time, using the low frequency image not only did not reduce the recognition rate of the algorithm and can greatly improve the efficiency of the algorithm, so that the overall algorithm has a good recognition effect and real-time effect.According of the computational complexity of the traditional feature extraction in the face recognition method and the characteristic of the traditional classification was too simple; this paper proposed a face recognition method based on 2DPCA, 2DLDA and fuzzy set theory. Through the use of 2DPCA and 2DLDA method greatly reduced the algorithm complexity, and improved the recognition accuracy; at the same time, the introduction of fuzzy sets has changed the deficiencies of the traditional recognition methods are too simple, and made the category decide more reasonable and effectively and further enhance the precision of the algorithm; and using the preprocessing method of discrete cosine transform not only can eliminate a large number of redundant information of the image, also can improve the recognition accuracy of the algorithm and greatly reduce the running time. The algorithm can behave more quickly and accurately.The improvement to the support vector machine technology of face recognition algorithm, this paper proposed two kinds of new recognition algorithm according to the different environment: one is face recognition algorithm based on combined kernel function SVM; the other is face recognition algorithm based on light discrimination and multi support vector machine. In the previous algorithm, the algorithm using the kernel PCA for feature extraction can make the face feature extraction more accurate and complete, and using the combined kernel function in SVM the new kernel function has better learning ability and generalization ability. This algorithm also using the preprocessing method of discrete cosine transform can make the algorithm shows better recognition effect and operation efficiency on the whole; in the latter algorthim, by light judgment in face image implement the image classification, this method can effectively eliminate the influence of light on the face image through wavelet transform in different scales image reconstruction method; at the same time, in the different scales using the multiple SVM identified the reconstructed image, finally getting the recognition results. In this paper, the algorithm has a very good recognition effect when the images were affected by illumination.Duing to the single feature extraction method of face recognition can not completely and accurately describe the face image; this paper proposed a face recognition algorithm based on multi feature fusion. This algorthm has improved the defect of original feature extraction method, at the same time fused several feature extraction mode. This algorithm has achieved more ideal extraction effect and recognition effect.The face recognition based on image is one of the hottest research topics in the field of computer pattern recognition and machine learning. It has a very high research value and broad application prospects. Because the current face recognition technology can not fully meet the needs of practical applications, there are still a lot of problems to be solved. In this paper, the difficulties in face recognition are studied and explored, and some research results have been obtained. Hopping can provide some reference and assistance to the scholars who study this technology in the future.