The Research On Human Detection And Tracking Of Intelligent Video Surveillance

Computer intelligent video surveillance system (CIVSS) is one of new arising high-tech application fields. The system can real-time detect, recognize, and track moving objects in a special environment. Furthermore, it can also analyse and judge the behavior of objects. The aims of CIVSS are to understand the meanings of video stream and to explain the scenes comprehensibly, hence to guide the action and make some decision.As to computer intelligent video surveillance, there are still many problems no matter in theory research or in applications. Large numbers of researchers devoted themselves in the area and have already achieved many progresses. The dissertation studied the key technologies of CIVSS based on the current conclusion. The main research of the paper is about the technologies about the human object detection, segmentation and tracking.Firstly, object detection technologies are investigated. Several object detection algorithms generally used are introduced, and the performance is analyzed. The self-adaptive threshold selecting is applied in the bi-directional projection histogram of grayscale two-frame differencing image during the segmentation moving region. The algorithm can effectively detect the information of moving human object from the video images.Secondly, the moving human object-localization technologies are introduced. Based on the video sequence captured from a fixed camera on the top of the scene and motion information silhouette of head information similar to a circle or arc in the detected region, a application of Hough transform is proposed in the paper. Multiple human heads detection is implemented using the Hough multiple circles' detection based on the continuity feature of silhouette curve.Then the human object tracking algorithm is investigated. The paper introduces the Mean Shift algorithm and its application in object tracking in detail. After the analysis of the advantage and disadvantage about Mean Shift, the paper combines Mean Shift algorithm with Kalman filter taking object moving and space color information into account and enhances the tracking performance.