A method for separating adhesive vehicles based on corner detection

In order to solve the problem of adhesive vehicles in video surveillance, this paper presents a simple segmentation method based on corner detection. Firstly, Shi-Tomasi corner detection is used to detect the corners of the vehicle area. Secondly, the corners are separated by K-means clustering method, and then the nearest corner from the line connecting the cluster centers will be found. And at last, the split line is determined with the nearest corner. With strong adaptability, the proposed method does not need any prior knowledge except for the vehicle shape. Experimental results show that this method is simple and effective, and it is potential to be developed for real-time.