Research On Detection And Segmentation Of Occluded Vehicles Based On Seven Grids

While using the video surveillance system to detect the traffic scene, the false detections often exist when more than one vehicle approaches each other. And this paper focuses on adhesions vehicle segmentation. Blocking background Image and detection image in the same way, a foreground image obtained by the method of Background difference. The use of shadow removal and eliminate noise points to deal with foreground image to calculate the movement area. Adhesion area of the vehicle is detected based on criteria adhesions vehicles criterion. Scan the vehicle contours with the "seven grids", to detect the recessed area between adhesion vehicles. By matching the corresponding recessed area, find the adhesion area between the vehicles. Using edge detection algorithm to detect the edge, regard the maximum amount of information as a suspected edge dividing line. Then, connect endpoint of the suspected line and midpoint of adhesion area, this segment is adhesions vehicle's dividing line, dividing the adhesion of the vehicle. At last, uses the vehicle tracking algorithm and the adhesion vehicle segmentation method together. And, the vehicle tracking algorithm is used to evaluate the accurate about the vehicle segmentation method.