A new cast shadow discrimination method for vechiles through edge detection

This paper presents a cast shadow discriminating method through the edge information of traffic images. In the process, we separate vehicle edges and shadow edges by analyzing their difference on the edge extraction image. In the paper, we will propose a neighbourhood detection algorithm and a line detection algorithm in locating the main curve of shadow edge. By the two algorithms the shadow edges are erased, and then the left vehicle ones are enhanced by a conditional dilation to reconstruct the vehicle. We apply result of the reconstruction to resolve the fake shadow problem, which is usually pitfall for traditional detection methods