Face Detection And Recognition Technology Research Of Video Monitoring System

In recent years, biometric technology has rapid development, and face recognition is the most widely used technology of all biometric face recognition technologies. Compared with conventional identification systems, the biometric identification system has a universal, unique, permanent feature, and in many places it has been recognized as a more secure identification system.In this paper, face recognition system, and a set of research programs are presented, specifically includes the following aspects.1. A research on the background and the development prospects of face recognition, as well as the detection and the recognition algorithms are systematically introduced. Face detection approaches are mainly based on knowledge or statical learning. Which are specifically described in the paper include PCA, LDA, DCT, SVD, wavelet transform, parameter model, elastic graphic matching, SVM, Neural Network, and 3D face recognition methods.2. System preprocessing, mainly includes video capture, image loading, body positioning, light compensation, median filtering, complexion modeling, gray projection horizentally or vertically, and face grabing. In which, the part of body positioning employed the background-difference method, which can quickly get a human target, and the complexion modeling can establish models for yellow and white skin, but less effective for the black skin, however, the gray projection method can achieve a great advantage for detecting black skin human targets.3. Face detection consists of three parts including eye positioning, mouth positioning, and face depicting. In the part of eye positioning, the sampled faces are sent to color matching and brightness matching first, then it receives the procedures of fake region cutting and expansion, and at last the centre spots of the eyes can be acquired. In the mouth positioning part, the sample is processed by brightness matching, color matching, pixel corrosion, discrete points deleting, fake region cutting, and the center of the mouth can be acquired as well as the eye spots too. If these three spot forms an acute triangle, the sample can be recognized as a human face.4. Face recognition on the images which have been thought as faces. First, the traditional PCA and Euclidean distance methods are used, but resulted in a low efficiency. So the wavelet transform to extract the characteristics of low-dimensional space is involved to be more efficient. In order to improve the recognition rate, the Gabor wavelet method and BP neural network are applied, and an obvious promotion is achieved. 5. Face recognition in 2D space can be more or less affected by bringtness, facial expressions and gestures. Therefore, it is necessary to expand face recognition from 2D space to 3D space. The phase decomposition of the grating projection image can lead to the generation of 3D face vector, which can be used for the reconstruction of human faces.This paper is completed by the home security systems research and design based on face recognition technology. The theoretical analysis, the system construction, computer simulation and experiments done by this paper can provide useful help for further research on face recognition technology.