Research And Verification Of Facial Recognition Algorithm Based On Kernel

Face recognition technology is a biometric identification technology to automatically identify by facial characteristics. Face recognition technology has many advantages in biometric identification technology, particularly intuitive, non-invasive, so it is used in a wide range of applications. For example, it can help the police identify sensitive characters (such as criminals, terrorists) and customs authentication. It can provide the authentication for all kinds of cardholder and be used for access control.A complete face recognition system include: face detection, face image preprocessing, feature extraction, classification and recognition. In the paper, two improved recognition algorithms are used to recognize. The focus for facial recognition in this paper is facial images pretreatment and facial recognition algorithms.Major content of this paper is as follows:1) The application of a margin classifier based on the kernel in the face recognition.For this kind of nonlinear problems of facial images, this paper proposes a novel large margin nonlinear discriminate analysis of kernel margin classifier. It combines the advantages of SVM and non-linear discriminate analysis. The samples after pretreatment is used to project the characteristic matrix into an implicit space called feature space by nonlinear kernel mapping, the kernel trick is used to improve the traditional large margin classifier algorithm, moreover, the theory of reproducing kernel in the new feature space is used to obtain the optimal kernel discriminate vectors with which the kernel within scatter is kept as small as possible. The method proposed in this thesis is experimented on ORL and Yale databases .The repeated experiment results show that the novel algorithm can increase the recognition rate comparing to the traditional SVM algorithm.2) The application of euclidean distance of images based on the kernel in face recognition.A new distance named euclidean distance in the image is proposed, which consists of the idea of kernel function. The paper design a novel recognition method named face recognition of euclidean distance of images based on the kernel . In order to verify the feasibility of the proposed method, firstly, it takes pretreatment to facial images to get pretreatment samples. Therefore, different dimensional sub-matrixes are extracted to compose a number of samples matrixes used to recognize (called DCT transform one); alternatively, primitive facial images are divided into 8*8 sub-block which is non-overlapping. Then, it takes pretreatment to every sub-block and different dimensional sub-matrixes are extracted to compose sample matrixes to recognize (called DCT transform two).The repeated experiment result show that the classifier with Euclidean distance in the image proposed in this paper has better recognition rate efficiency than other distance classifiers.