The Research Of Human Movement And Identity Recognition In Surveillance Video

With the rapid development of computer technology and the improvement of living standard, the requirement of security in many areas is getting higher and higher. For example, to monitor the residential district intelligently by computer technology. It is not only saving the manpower resources and cost, but also providing evidence for future investigations. Intelligent surveillance is to detect human movement object automatically by computer without judging by people. The goal of intelligent surveillance is to detect the type of human movement and identify who the moving object is; the current research level has a long distance from the goal.In this paper, we research on the ideal surveillance video according by the current research level, they are the human movement recognition of surveillance video with static background and color iris recognition, the iris image used to recognition only contain iris and pupil. The study of the two ideal cases is useful for the future of the biological feature recognition and the intelligent control. The content of human movement recognition are, to extract the background of video adaptively and real-time, segment the human movement object from the video, extract the SIFT feature of human movement object and identify the type of movement by SVM. The iris recognition is mainly to extract iris color information descriptor with CBSILF and match the iris image tested with the irises in the database.There are three traditional background extraction methods of monitoring video. One method is to take a photograph as the background of video, but the object segmented is easily influenced by the light, weather and so on; the second method is to average the values of the corresponding pixels of the video image sequences, and put the result as the background of the video, but this method may cause noise in the segmented object; the last method is to extract the video background adaptively, though many scholars do research in this field, the result is still not good. Compared with the traditional background extraction methods, the proposed method can real-time extract the video background adaptively, there is almost no noise in the segmented object and the SIFT features extracted can well reflect the movements. The experiment shows that the recognition rate of the proposed method is95%. Most of the current iris researches are basing on the gray image, even if some scholars research iris with color image, it is also converted to gray image. This paper extracts the iris color information descriptor, the feature information can well reflect the iris’s character, and the recognition rate is100%.