Motion Detecting And Tracking Technology Based On Video

With the continuous expansion of the scale of video surveillance, the traditional way which only rely on human to monitor the thousands of online camera and to analysis the numerous surveillance video has been unable to meet the demand. The intelligent surveillance system does not require human’s efforts and can automatic analysis the sequence captured cameras. Therefore, the intelligent surveillance system becomes the future trend.The article mainly discuss the motion detecting and object tracking technology. Using the practical application in Intelligent surveillance to further explain the technology. On the one hand, intelligent analysis of recorded video from surveillance system, extracting an object of interest. On the other hand, automatically control camera for real-time tracking of the target and expand the scope of monitoring.First, the article introduces all kinds of background modeling algorithm. Discuss some common skills and test them. I provide more information about the shortcoming of Gaussian mixture model algorithm. On the basis of the original algorithm, I fuzzy the spatial correlation to improve detection accuracy.Second, the article discusses the target tracking algorithm. We introduce several typical algorithms, like Kalman filter, LK optical flow method, Camshift, Tracking-Learning-Detection method and so on. We did some test and give experimental analysis.Then, the paper describes the specific application of pedestrian-vehicle detection and classification based on the video. The application achieved through five modules: Image Preprocessing, Motion Detection, Tracking, Classification and Learning Patterns of Motion.In Classification stage,we present a method that treat target area as feature and use EM clustering as classifier. Model estimation phase, using statistical history of the scene to fit the different position of the target area. Finally, depending on the application needs, given the actual surveillance video of the test results, data and analysis.Finally, the paper describes the automatic camera tracking application. This application contains three modules:Motion Detection, Tracking and Camera Control. The article then gives the experiment result of human face and real-scene.