Research On Moving Objects Detection Throw From Vehicle Based On The Video

In recent years,with the rapidly development of the computer science, image processing, network technology and transmission technology, intelligent video surveillance becomes a hot research filed of computer vision topics. Intelligent video surveillance can identify and judge moving object without human intervention, the warning can be issued in unusual circumstances, it is effective to assist security personnel to deal with risk timely, and possible criminal acts are avoided. As one of the important technologies to achieve intelligent video surveillance, target detection and tracking based on motion video attracts large numbers of researchers devoted themselves in the area and many progresses have been achieved already. Based on these previous studes, By using the distance between object and Vehicle, judges whether there is object being thrown out, and combining with the frame difference method, a method based on the biggest probability value supposition background restructuring is improved. Main content can be summarized as follows:1. The ordinary techniques of moving target detection and object segmentation are briefly introduced, and the advantages and disadvantages of each method is analysised. An algorithm based on Gaussian Mixture Models(GMM) is used to detect small moving objects, the algorithm can adapyively update the background, and the interference caused by the leaves tremble and the background change is avoided.2. The background reconstruction algorithm based on maximum porbability assumptions is researched deeply. This single method dose not work perfectly, when assumptions dose not met, the background reconstruction can't be accurate, but this problem can be solved by combining with frame difference algorithm. In addition, a motion detect method based on the RGB color model is introduced, which combines area filling method with dilate method in the binary image processing. And the deformation of target contour caused by excessive dilates can be minimized.3. According to the law of the distance between vehicle and camera which are related with the size of the vehicle in the video, and by the use of the distance and the area ratio, the problem whether there is object thrown out of vehicle can be solved. And the mean shift algorithm is used to tracking vehicle, red mark is used to tracking the vehicle when the vehicle thrown object. The results show that it works effectively. At the end, the system interface is designed, which makes it is more intuitive to detect the process in which object thrown out of vehicle.