基于扩展数学形态学的车牌定位算法

An Algorithm of License Plate Location Based on Extended Mathematical Morphology

在汽车牌照识别系统中,车牌定位是整个识别模块实现的前提,目前车牌定位的方法多种多样,各有所长,但存在着计算量大或定位准确率不高等问题。边缘检测是常用的车牌定位方法,边缘检测的质量决定了车牌图像的最终定位结果。一般人们习惯于用基于梯度和基于模板的算子提取边缘,但这类算子都不能很好地滤除噪声,因而给噪声图像边缘检测带来了困难。根据数学形态学原理与方法,提出一种扩展数学形态学车牌图像边缘检测算子,并结合水平和垂直投影进行车牌定位。实验结果表明,该算法不仅能成功提取车牌图像边缘,而且能很好地滤除噪声,从而实现准确车牌定位。

In car license plate recognition systems, license plate location is the precondition of the whole recognition module. Now various methods are used in it, each of which has its own advantage. However, such problems as the quantity in calculation or the low correct location rate are not solved. Edge detection is a common method of license plate location, but edge detection＇s quality decides the result of car license image locating finally. Generally, people are accustomed to de tecting edges by using gradient-based algorithms and template-based algorithms, but all of them fail to filter the noise. Ac cording to the basic theories and methods of mathematical morphology, this paper puts forward a car license image edge detection operator based on extended mathematical morphology, both the horizontal projection and vertical projection are used to locate the license plate area. The experimental results indicate that the novel proposed algorithm is a better method for car license image edge detecting and noise filtering than the usually used ones, thus realizes the accurate locating of car license plates.