License plate detection in complex scenes based on fusion of Gaussian filtering and Bayesian network

In this paper, the problem of automatic license plate recognition is addressed. We focus on detecting the location of the license plate, which is one of the most challenging parts of such systems. Due to the strength of edges within the plate region, most existing methods start by edge detection. However, relying only on edge detection is subject to failure in complex backgrounds and bad-illuminated images. To soften these conditions, we propose to fuse two strong tools, i.e. Gaussian filter and Bayesian network after preprocessing and edge detection, to achieve acceptable detection rate. Our experimental results confirm the effectiveness of the proposed method in detecting the license plate in low/high illumination conditions and complex backgrounds.