An adaptive approach to vehicle license plate localization

It is well known that locating vehicle license plate is a very difficult problem in the field of image segmentation and ITS (intelligent transportation system). And unfortunately most of the methods already presented don't work robustly and usually impose some special restrictions on the input image condition, such as illumination, noise, etc. In this paper we proposed an accurate and robust localization algorithm by employing an adaptive thresholding and search strategy. The algorithm is based on the spatial frequency characteristic that the grayscale in the license plate region usually varies more frequently than in the background due to the existence of characters. In this paper, an adaptive threshold is firstly reached through an iterative selection-verification process. The gray image is then binarized by this threshold, and the algorithm localizes the license plate in the binarized image combining some prior knowledge of the license plate. The final accurate license plate segmentation is achieved after an accurate edge detection process based on the color information of the license plate. The robustness and performance of the algorithm are demonstrated by the convincing segmentation results even when the images are rather complex, noisy and in poor illumination.