Research And Development Of Intelligent Video Surveillance System

Intelligent Video Surveillance System is an advanced field of research, which involves image processing, pattern recognition, artificial intelligence and computer vision. What is more, intelligent video surveillance system plays an importance role in both security domain and non-security domain. Due to the commercial value and application prospect, intelligent video surveillance system has generated great interest at home and abroad.The thesis mainly aims to study the intelligent video surveillance system. At the same time, the thesis research on the current status of the intelligent video surveillance system at home and abroad and the problems which puzzle the researchers at present.In the intelligent video surveillance system, many intelligent objects are based on moving object detection, so the accuracy of the detection determines the accuracy of the intelligent video surveillance system. At first, This thesis analyse the advantages and disadvantages of the common moving object detection techniques, such as background subtraction, frame differential method and optical flow, and investigate some improved algorithms of these techniques. Since optical flow does not meet the real-time requirements of intelligent video surveillance system, intrusion detection technique based on videos was investigated according to background subtraction and frame differential method. Additionally, this thesis proposed a method to automatically detect moving object intruding into security area and give an alarm, according to the relationship of artificial security area and moving object.As an intelligent video surveillance system, the evidence after an accident is very important. The thesis includes time-based video retrieval, event-based video retrieval and content-based video retrieval. The time-based video retrieval finds video according to the time which user inputs. The event-based video retrieval finds video according to the intrusion detection results. In order to implement video retrieval based on video contents, unstructured video sequence should firstly be structured to scene, shot and frame, then to analyse video structure (boundary detection of the scene, and extract key frame from well adjusted scene). This thesis introduced some scene boundary detection and key frame extraction techniques.Finally, the intelligent video surveillance system is designed with Microsoft Visual Studio2005and OpenCV2.0. The system has some basic functions which most video surveillance systems have, for example, displays of multi-channel video, PTZ settings and control, Internet video transmission and so on, but also some smart functions such as intrusion detection of security area and content-based video retrieval.