The Research And Realization On Face Recognition Algorithm

Face recognition technology caught research attention in the early 20th century and it had always been a hot related research topic in image processing, machine vision and pattern recognition field. As an important branch of pattern recognition, face recognition plays an important role in practical applications such as public security survilance, bank ATM transactions, network security, mobile communication and visitor monitoring system. Face recognition involves the multiple classification problems, which has a great research content.In this thesis, based on some related fundamental theories of face recognition technology, the curvelet transform and principal component analysis, we have carried out some research investigation to take the advantage of each algorithm. The research work and thesis innovation are as follows.1. Since collected face images were influenced by illumination, posture, facial shelter, expression change, we adapted the face image preprocessing methods in this paper, including gray level transformation, histogram equalization, median filtering, image normalization and so on. Also, we have obtained the precise location of human eyes in order to extract the valid district of human face.2. KPCA algorithm can solve the problem of nonlinear characteristic that the PCA algorithm can’t handle with. So we put forward KPCA algorithm. The KPCA algorithm has a good effect on extracting face contour and the curve detail information through internal nonlinear kernel function. Support Vector Machine (SVM) has the strong ability of classification of small samples and the advantage of dealing with nonlinear and high dimension. In this paper, the KPCA and SVM methods were combined for face recognition. At first, we obtaine the low-frequency component of the face images decomposed by curvelet transform, then the feature vectors were extracted by KPCA, and the strategy of "one vs one" of SVM was chosen to perform recognition.3. The traditional curvelet decomposition algorithm cannot take full advantage of the fine scale composition information, so we put forward the data fusion algorithm through analysis. Data fusion algorithm can make full use of face feature information of each dimension of curvelet decomposition algorithm through combing with the different scale component decomposed by curvelt decomposition according to certain proportion. Then the curvelet faces dimension were reduced using PCA. The coarse information and the fine information were combined by data fusion. Fnally the nearest neighbor algorithm was used for classification.