Research On Detection And Tracking Of Moving Object In Intelligent Video Surveillance System

Intelligent Video Surveillance System is a process that automatically analyzes image sequence to achieve target detection and tracking in the dynamic scene by using video analysis without people. On this basis, some follow-up steps can also be achieved, such as target recognition, target behavior analysis, alarm and so on. Thus, intelligent video surveillance technology is one of current hot issues of domestic and international research of computer vision.Moving object detection and tracking is important parts of video monitor system, so how to detect and track moving object steadily, real-time and effectual, becomes an important question that needs to be paid attention and researched. The dissertation aims to design intelligent video surveillance system with static cameras. Based on studying the current conclusion, the dissertation improved and realized the key technologies which relate to detection and tracking of moving object. The major contents of this dissertation can be summarized as follows:In the area of target detection, some common methods of moving target detection were analyzed, and the method of background subtraction based on Gaussian mixture model was chosen. Then the problem of setting the threshold manually was analyzed, which is necessary when targets are determined in traditional Gaussian mixture models. And this problem is in the way of improving practicality of moving target detection system. To solve this problem, a Gaussian mixture model based on adaptive threshold was proposed in this dissertation. On the basis of analyzing several adaptive threshold, traditional Gaussian mixture model was improved with using the OSTU method instead of the method of detection with fixed threshold, which make moving object detection system could dynamically set the threshold according to scenes. Therefore, practicality and accuracy of moving object detection system were improved.In the area of target tracking, CamShift algorithm was chosen to accomplish the task of tracking. To solve the problem that the effect of tracking fast moving target with CamShift algorithm was unsatisfactory, the traditional algorithm was improved by predicting the target location in next frame of video with Kalman filter. Improved CamShift algorithm got an iteration starting point which is closer to real location of target. Therefore, searched area was smaller, and iteration times were less. So, tracking speed and tracking precision was improved, and reliability and robustness of object tracking system was improved too.The results of tests show that the algorithms in this dissertation are effective, which can be adopted in actual scenes to detect and track moving object.