The Research Of Multi-face Recognition Algorithm For Intelligent Video Surveillance

Intelligent video surveillance system (IVSS) is an advanced intelligent video analysis system with the function of intelligent analysis, recognition and processing on images. Face recognition is not only a hot research topic in biological recognition technology, but also one of the most difficult problems in the field of artificial intelligence. With the deep development of peaceful city and community, it put forwards higher request for the city intelligent surveillance system, especially the face recognition technology in intelligent video surveillance gains extensive market demand. The eternal law is as follows that market leads technological development. In view of the market demand, this paper expands the face recognition technology in intelligent video surveillance and obtains the new algorithm to gain real-time detection and recognition for multi-face of video image after reading many relevant literatures. The main research work is as follows:1. The development and application for the technology of face recognition and intelligent video surveillance system is integrated respectively, and the basic theory related to data processing such as multi-core parallel programming theory and image preprocessing method is summarized.2. To improve the video image multi-face parallel detection algorithm, the development, the basic concept and principle of Adaboost algorithm are introduced; the process of face detection in video surveillance image is analyzed and explored in detail,and then add on an additional detection region. The results show that the refined algorithm can markedly improve the rate of correct detection and reduce the time of face detection to a certain extent, so it meets the real-time requirement for video-image detection.3. During the processing on PCA in face recognition, this paper gives a PCA algorithm for multi-face recognition based on f-k optimized nearest neighbor method by optimizing the minimum distance among different projection coefficients of characteristic space. The new algorithm makes full use of indirect information among consecutive frames of the video, improves the rate of correct recognition of face recognition under certain conditions and strengthens the warning capability of classification of suspicious persons..4. To carry out the integration test for the improved PCA face recognition algorithm and multi-face parallel detection algorithm, the test system for intelligent video surveillance multi-face recognition is built in VS2010software platform. The parallel multi-face recognition results show that the new algorithm get better application in the test system than the referenced one, and realizes the purpose of real-time multi-face detection and recognition in the video.