The Research Of Face Recognition Under Unlimited Conditions

Along with the development of informative process, digitization and network, the safety of national as well as social activities is faced with new challenge. Under this kind of environment, traditional safe technology shows a lot of abuses and defects. Biological feature certification technology is the one of better method of solving the safe problem of this new challenge. Face recognition has its specific advantage, may be extensively applied in the department of public security, convict identify search, punishment detective case, safe verification and supervisory system, the human-computer interaction control, certificate check, passageway control and verification of credit card.In the thesis, the fundamental theories and the key technologies of face recognition under unlimited conditions are mainly researched. The topics such as preprocessing of face images, face feature extraction, classifier design are discussed. Under the unlimited conditions, preprocessing of face images is a very important step in face recognition. The geometry correction and the handling of gray scale and shelter of the face images have been discussed in details. Preprocessing of face image have been proved useful for raising the rate of face recognition by contrast experiments. In the part of face feature extract, the thesis give a detail discussion of Eigenface recognition, present a improved method for Eigenface recognition, show its well recognition result through contrast experiments. At the same time, the article introduces the application in two-class classification of the support vector machines first, then, the methods of multi-class classification of SVM are presented. Face database is important for face recognition, in the part of design of face database, the article provides a fast and easily method for face database design based on delphi and make a sample based on access and ORL face database. With the interaction of delphi and matlab ,a efficient face recognition system combined with the PCA and SVM has been realized.The results of experiments show that preprocessing of face images has important influence for improving the face recognition rate. Making different Eigenface space for variant pose is the effective method for solving the face recognition of variant pose on many test multi-pose samples.Compared with Euclidean distance classifications, SVM has its special advantage on small samples classification problems.A general face database may be helpful to rise the efficiency of face recognition and promote its develops to network direction.Combined with the delphi and matlab can validate the face recognition algorithms conveniently.