A license plate recognition algorithm under low illumination environment

Abstract

This paper presents a solution for the license plate recognition problem under low illumination on the basis of in-depth analysis of the characteristic of license plate images. In recent years, with the development of technologies such as optics, computers and pattern recognition, license plate recognition technology is improved constantly, but the accuracy of license plate recognition under low illumination environment still needs to be improved. In this paper, we employs remarkable texture feature from license plate area to find the approximate area of the license plate. Then we determine the final location of the license plate area according to the candidate's gray distribution features. The proposed localization algorithm is superior to the traditional background subtraction method. After obtaining the license plate, this paper segments characters through skew correction, image enhancement, projection segmentation, and prior knowledge. Finally, two kinds of algorithms are used to identify Chinese and alphanumeric characters respectively, because of the differences in the structure of Chinese characters and alphanumeric characters. The algorithm is tested with 207 vehicle images under low illumination. Experimental results show that the accuracy of proposed algorithm for license plate localization is 98.06%, the accuracy of character recognition is 94.90%.