Research On Identity Distinguishing System Based On Real-Time Face Recognition

Face recognition (FR) is one kind of newly developed biometric identification technique, which recognizes persons based on face images. Compared with other technologies, such as fingerprint recognition, iris recognition and so on, FR could figure out people's identification without special cooperation, which is an excellent advantage. Therefore, it interests many people once announced. The related fields of FR includes: image processing, pattern recognition, computer vision, artificial intelligence and so on.It is introduced in this paper that how to work out a real-time identity distinguishing system based on FR. Therefore, our paper involves all technologies related with FR, the main points of which are listed below.1,Face detection (FD). FD is the first step of FR. In this paper, firstly, early developed detection technologies are introduced; secondly, FD algorithms based on skin color are discussed, which have been widely researched during the past time; at last, well-known cascade based on AdaBoost algorithm which is used in our system is presented. It is good at real-time detection, which is just what we want to for real-time processing.2,Face recognition (FR). FR consists of two main parts, namely, facial feature extraction and classifier design, and the former is much more important than the latter, which is the critical step of the whole system. In this paper, firstly, widely used principle component analysis (PCA) is introduced; secondly, linear discriminant analysis (LDA) is presented, which emphasizes within-class and between-class, especially; at last, our own designed global search PCA and LDA are presented, which are aimed at improving the weaknesses of PCA and LDA, respectively. The former could find better class-related feature space, and the latter help LDA to solve small sample size problem. Those strategies have been tested on several standard face databases. And then, classifiers based on distance functions and neural network are discussed. In our system, a classifier based on RBF neural network is used.3,System integration. In this part, how to integrate a FR system is introduced, and our own system is demonstrated, which is good at real-time and accurate recognition. This system could identity several persons in one picture.