| With the development of society, people pay more attention to the security. A fast and effective identity verifying method is strongly demanded. Face recognition, in a simple word, is using the method of computer vision and pattern recognition techniques, extracting the features of face images, then identifying the person in the image. The technology of face recognition is an active subject in the area of pattern recognition, and has broad applications in the field of law, business, etc.Face recognition mainly include three parts: l.Face detection and pretreatment; 2.Feature selection and extraction; 3. Classifier design, and recognizing faces by it. In this paper, we mainly discuss some problems in image pretreatment and feature extraction. The work includes:1. Facial image pretreatment: Eliminate the impact of background, hair, clothes, etc, and keep the main information need by recognition, which is propitious for extracting features and recognizing face under the same condition. A method is described for locating eyes accurately in this paper. It combines the integral projection method and the differential projection which is proposed in the paper. So that it provides a datum mark to the face rotation and size normalization.2. Choose the appropriate feature extraction method which should both keep the tiny difference of different faces and the clustering of the same persons faces. In this paper, two classical algorithms that based on the algebraic feature of image, PCA and LDA, are introduced. And a method based on the between-class scatter matrix under the 2DPCA algorithm is put forward. The 2DPCA algorithm was put forwarded in 2004.Then we use the nearest-neighbor distance classifier to recognize and get a relatively high recognition ratio.3. Transplant and optimize the codes on DSP. The chip DM642 which belongs to the TI TMS320C6x series chips is chosen. And we use the C language and the assembly language to optimize the codes on DSP.
|
PDF Full Text Request