A New Method for License Plate Detection Based on Color and Edge Information of Lab Space

Abstract

License plate recognition (LPR) is one of the key technologies towards intelligent transportation system. Whether license plate can be detected precisely or not may affect the whole system's efficiency. Existing algorithms based on gray image have a poor result when a complicated background or a blurred license plate occurs, while algorithms based on color space cannot deal with the situation where the car and the plate have the same color. In order to solve these problems, a new method based on Lab space is presented in the paper. Firstly an algorithm for determining plates' color is developed. Then for blue-plate-included images and yellow-plate-included images, threshold matrix is selected respectively, which offers possible optimal thresholds in turn. And then the true plate area is filtered in accordance with the license plate prior knowledge. To tackle the problem that the car and the plate have the same color, a method based on vertical edge analysis and character region mergence is proposed, which can realize the secondary license plate location for such images. Experiments show that this algorithm can overcome the limitations of both the color-based methods and the gray- based methods and is less time consuming.