An Effective License-Plate Detection Method for Overexposure and Complex Vehicle Images

License plate detection plays an important role in license plate recognition (LPR) system. If license plate can be detected exactly, the character segmentation and recognition can be implemented more precisely and efficiently. In our observation, the illumination affects the result of detecting the license plate. The aim of this paper is to extract more than one license plate from an image and to obtain the license plates in the image which is taken under different conditions. The proposed method includes six processes as follows: transferring color to grayscale, image equalization, edge detection, checking black pixel ratio, plate verification and output license plates. The experimental results show that the proposed method can gain 94% accuracy in GetRatio and 83% accuracy in GetRight.