Research On Household Anti-theft System Based On Face Recognition Technology

As a new surveillance technique, video surveillance system has been widely used in many fields in recent years. The human body biometric identification technology into the video surveillance system is a new direction for the development of Video Surveillance System. A household security video surveillance system which is used face recognition technology has been designed in this paper.In order to adapt the change of surveillance environment, the system is designed for gray-scale images. System consists of three parts, according to data flow in the order, respectively video capture module, face detection and localization module and face recognition module. In this paper, we adopt Microsoft's multimedia software development kits --DirectShow structures in the framework of the whole system,at the meanwhile, the kits can be also used to accomplish the assignment of video capturing without the aid of video capture card, which not only facilitates the operation of the system, but lowers the costs. As for the aspect of the face detection and locating, according to the features of this video surveillance system, a method of comparing the difference between present frame and system background has been adopted to get body religion. And a strategy of system background institution has also been designed, then a board geometric projection can be applied to the given body region, and the corresponding candidates will be made conclusion. Finally, face locating can be performed accurately with the guidance of the mosaic concept. When it comes to the aspect of face recognition, use the principal component analysis in order to project high-dimensional face image to relative low-dimensional space, which realizes the goal of decreasing dimension. Then use linear discriminant analysis method to make the vector given by last step projection to the best classification space. The coefficients, which second projection has reached, are just features of face image. Finally a three layer-BP neural network classifiers can be used to achieve the classification of face features, that is to say face recognition. As for family anti-theft system, when a fresh face has been recognized, it means that there are invaders appearing.The experimental images are all collected by Camera, including single face, muti-faces and different face expression. Experimental results demonstrates that as for a small image library ,including 105 individuals face, each size 320×240 pixels, face detection and locating of the detection module has reached to 93 true face, 5 false face. The average detection time is 600ms on the Pentium IV, the PC memory 256M PC. For 20 categories of different faces, each 6 pieces, a total of 120, each small face size is 64×64 pixels, and face recognition module recognition correctness rate has reached to 95%.And average recognition during time is approximately 150ms in laboratory. The accuracy and time-consuming performance of these two functional modules can meet general requirements of the surveillance system.