The Research Of Face Recognition Based On AdaBoost Algorithm And Fisher Criterion

As a long-term concern of the field of pattern recognition, face recognition of machine vision technology is an important topic, with a high value of academic research and commercial applications. As an important aspect of biometric identification, face recognition has broad application prospects in the file management system, security, authentication systems, credit card verification, the public security system criminals, identification, banking and customs surveillance, human-computer interaction, and other fields. Face recognition system is divided into two parts ,first of all, detect the existence of face images from one of the scenes of the static or dynamic video images, if the existence of a number of known storage identity Face Image database verification and identification of a single scene or personal identity. The non-rigid facial features ,vulnerable to light and the effects of posture make face recognition facing great challenges.In this paper, the face detection and face recognition are researched partly.First based on the research for the face detection algorithm with AdaBoost algorithm. Analysing the convergence of AdaBoost algorithm performance, and the weight of generalization ability to update methods of the classification of the impact of the in-depth analyse. On this basis, a new method of updating a sample of parameters is advanced, the method has changed the traditional AdaBoost algorithm weight adjustment process, and more concerned about the correct sample weights amendments to improve the detection efficiency.In the traditional process, the global features is used as image classification and identification information for face recognition, in the various methods ,concluding the convergence of principal component analysis and linear discriminant principle of the method of identification . Fisherface is best. The study shows that global features and local features are indispensable for face recognition. Against this, this paper presents the characteristics and the overall integration of the local features Fisherface method. First of all, parts of the local characteristics is evaluated by the distance weight, the right size reflects the value of all identifiable effects on the local characteristics of the contribution. Because of the limitations of with local and global features partly in face recognition. The Fisherface method based on the fusion of global and local facial features is presented .Fuzzy again based on the criteria for integration of the two methods on the identification result, the new algorithm can be very good advantage of the integration of the two to make up for the lack of both, thus enhancing the recognition rate.