基于数学形态学及车牌综合特征的车牌定位算法

摘要

The location of license plate plays a critical step in license plate recognition system. A practical and effective method of locating the plate was introduced by using the rich edge information and unique features of the license plate. Firstly, the original image was transferred to gray space and the rough location were detected by using the edge information of characters and the operation of the mathematical morphology and accurate region of license plate was decided by the Bayes classifier using five features proposed in this paper. Finally the experiments on 1500 color images show the efficiency of the proposed algorithm is 95. 20%.%车牌定位是车牌识别系统中的关键步骤;利用车牌图像区域丰富的边缘信息以及车牌本身的特征,提出一种实用而有效的车牌定位方法;首先,将原始图像转换到灰度空间上,利用车牌丰富的图像边缘特征信息和数学形态学操作对图像进行粗定位；然后,根据车牌本身的特征量化5种不同的特征,通过贝叶斯分类器的训练,实现对车牌区域的精确定位；最后,通过实验对1500幅彩色图像进行测试,其有效率可达95.20％.