Abstract: With the development of intelligent transportation, license plate recognition system has become an important part of it. License plate recognition can be divided into three procedures, including license plate location, character segmentation and character recognition. In order to achieve accurate license plate recognition, a novel approach is proposed in this paper. During the task of locating license plate, the double-edge detection method is adopted for positioning the license plate. And the combination of finding connected domains and traditional?segmentation of projection is applied for character segmentation. Regarding character recognition, three kinds of classifiers are utilized for improving classification accuracy and the strategy of reclassification of confusing characters is employed, which can shorten the training time and get higher accuracy. Experimental results demonstrate that proposed approach is able to achieve high recognition rate with reasonable computational complexity.

Key words: license plate location, edge detection, character segmentation, connected domains, vertical projection, character recognition, Support Vector Machine（SVM）

摘要： 随着智能交通的不断发展，车牌识别系统已经成为其中的重要组成部分。车牌识别分为车牌定位、字符分割以及字符识别三个部分。提出了一种新型车牌识别方法。在车牌定位方面，采用双边缘检测车牌定位方法；对于字符分割则提出了寻找连通域与传统投影分割相结合的方法；在字符识别上，将分类器分为三组，同时对于易混淆的字符进行了再次分类，这种做法缩短了训练时间，提高了准确率。实验结果表明，所提出的方法具有识别率高和速度快等特点。

关键词: 车牌定位, 边缘检测, 字符分割, 连通域, 垂直投影, 字符识别, 支持向量机（SVM）