基于MSER与边缘投影的车牌定位算法

摘要

针对基于颜色或边缘的车牌定位算法常出现误定位的问题, 提出一种适用于嵌入式平台的基于MSER与边缘投影相结合的车牌定位算法.对RGB彩色图像进行灰度变换突出车牌区域, 使用MSER算法进行连通域分析获取车牌位置;对于车牌和车身颜色一致的情况, 设计垂直边缘投影算法进行补充定位;利用车牌的字符间距特征进一步提高车牌定位的准确度.实验结果表明, 该算法对车牌倾斜、弯折、光照不均、大小不一、背景杂乱、分辨率低等情况具有很好的鲁棒性, 准确率可达97.1％, 适合应用于嵌入式车牌识别系统.%To solve the problem of current license plate location algorithms based on color or edge usually get false positive candidate regions, a method of license plate location which combined MSER feature and edge projection algorithm was presented.Gray level transformation was applied on RGB color images to highlight license plate regions, MSER algorithm was used to perform connected component analysis to obtain the position of license plates.For the case of the color of the license plate being the same as the color of the car, vertical edge projection algorithm was designed for complementary.By using the character spacing feature, the location accuracy was further improved.Experimental results show that the proposed method is robust to license tilt, buckling, complex background and low resolution with 97.1％accuracy and it is suitable to be applied to embedded license plate recognition system.