摘要
车牌定位与字符分割是车牌识别系统进行字符识别前重要的两个步骤,论文将介绍一种高效的基于最大稳定极值区域(MSER)的车牌定位与分割算法。首先对图像进行预处理并提取MSER,根据MSER间几何关系将相邻的MSER聚类在一起作为一个车牌候选区域,再利用机器学习及标准车牌的特点对每个候选区域进行分析,定位出车牌区域。然后将车牌区域根据字符的个数及MSER间关系划分为不同等级,并对不同等级的车牌采用不同的分割算法。实验数据表明,该方法车牌定位的准确率是99.07%,字符分割的准确率为97.9%。
License plate location and character segmentation are two important steps towards character recognition of vehicle license plate recognition system.In this paper,an effective algorithm is introduced based on MSER.The algorithm preprocesses the image and extracts MSER,then determines a candidate area of license plate by the clustering based on geometric relation of MSER.After the process,the license plate area is located by analyzing each candidate area based on Machine Learning(ML)and then classifyed into different ranks by number of characters and relation of MSER.The license plate area uses different algorithms of character segmentation for different ranks.The experimental results show that location precision is 99.07% and character segmentation precision is 97.9%.
出处
《计算机与数字工程》
2015年第12期2271-2274,2294,共5页
Computer & Digital Engineering
关键词
最大稳定极值区域
车牌定位
字符分割
maximally stable extremal regions
license plate location
character segmentation