Research on Some Techniques of Intelligent Video Surveillance
|Course||Applied Computer Technology|
|Keywords||intelligent video surveillance abnormal behavior detection codebook model crowd density estimation crowd motion estimation|
With the development of economics, public places face with many serious security problems. How to manage and control the crowd effectively comes to be an important issue which we have consider nowadays. The emergence of intelligent video surveillance has resolved this requirement to a certain extent, and intelligent video surveillance technology comes to be a newly arisen research direction. Thesis aims to extract the information of behavior characteristics on the video surveillance by the computer vision technology, image processing technology and artificial technology.The information is feedbacked to the terminal. Finally relevant departments will deal with the abnormal event. Intelligent video surveillance technology has a broad prospect and potential economic value in bank, electric power, transportation, post and telecommunication and other areas of applications, while saving a lot of investment in human and material resources. It becomes a hot research area.Actual scenes in intelligent video surveillance is often very complex and volatile, so object detection is particularly important, It directly affects the follow-up procedures. This paper is committed to detect and identify of people abnormal behavior which including crowd’s density and motion estimation under the specific surveillance scenes. First of all, the development of intelligent video surveillance and research status is studeied. Then the main methods of moving object detection in the intelligent video surveillance is explored, we compared these detection methods by experiments. In order to meet the requirements of real-time intelligent video surveillance, a improved codebook method was used in this paper for foundation of detection method.In terms of population density estimation, a linear function was obtained by fitting analysis of statistical characteristics and crowd density, and then it was used to estimate the crowd density. For crowd motion estimation, a block match algorithm is adopted. With experiments, we compared the Full Search algorithm and Three Step Search algorithm on both superior and inferior. At last this paper proposed to use Three Step Search algorithm in intelligent video surveillance.