Segmentation method of moving vehicles based on semi-fuzzy cluster

Moving vehicles segmentation is the most fundamental and vital problem in the intelligent transportation systems. This paper proposes a moving vehicles segmentation method that combines the semi-fuzzy cluster algorithm under the guidance of edge-based information with the traditional background subtraction algorithm. Above all, the current frame image that involves moving vehicles is divided into two parts using the algorithm of edge detection and edge closing. One part is the set of edge pixels and the other one is the regions encircled by the edge pixels. And then, every edge pixel will be associated into the most reasonable region according to the semi-fuzzy cluster algorithm. At last, the regions that similar with the background will be deleted and the remained regions are the moving vehicles in the current frame. Simulation experiments show that the new method posed in this paper is more robust and exact, and have a high ability of anti noise together with a high application value.