The Design And Implementation Of Static Background Multi-target Detection And Tracking

Ship on the water monitoring technology in our country at present stage in the development of very quickly, on the inland transportation is part of China's economic development. However, in the actual voyage many unexpected happens, such as ships collision, the ship on the rocks, and between the ship bridge collision, these are to human life safety and property security caused great impact. So, for the development of the detection and tracking of targets on the sea is very necessary, it can not only help reduce accidents, but also can quickly after the accident. Intelligent monitoring system is an important project of our country present stage research. For maritime control system development is a difficult and complex topic. In recent years, there is a lot of scholars and experts to target detection and tracking is studied on the sea.In this paper, by comparing the background difference, frame difference and optical flow method of these commonly used detection algorithm, because this is the background static experiments, use cases, so finally chose the background difference method in the Codebook background modeling method, through to each pixel to set up a code, through the establishment of code this background image, and then through the foreground image and the background image difference is able to detect the movement of the target very well. When motion target into the video will give movement target number, and extract the moving object data information, the information stored in a text document. So it can detect the moving target very well. This article USES the camshift, meanshift is it continue to use. It is will each frame of the search window size and center behind as a frame of the initial value to the target tracking. It can effectively solve the target deformation and occlusion problem. Under the background of simple can get very good tracking effect. There is a little complicated to join the prediction algorithm. This article to join kalman filter to track prediction of moving targets, when the target overlap when separated from effective tracking of the original goals. Through the video information to predict the next frame image motion target location. So when multiple target block occurs, through kalman filter to predict trajectory, when moving object apart, can according to the previously stored data information of moving targets, can accurately determine the original goals.In this paper, the OpenCV digital image processing for learning, for later research on multi-target detection and tracking algorithm has laid a solid foundation. Based on OpenCV and VS2010 moving targets for video image analysis has good real-time performance, to the right of multi-objective real-time detection and tracking, and has good robustness.