一种基于边缘信息的车牌图像分割方法  

A METHOD OF IMAGE FOR LICENSE SEGMENTATION BASED ON EDGE INFORMATION

作　　者：王戈壮[1] 张太镒[1] 游飞[1] 

机构地区：[1]西安交通大学电子与信息工程学院,陕西西安710049

出　　处：《计算机应用与软件》2008年第12期187-188,217,共3页Computer Applications and Software

摘　　要：为提高车牌预处理过程中字符与背景的分隔效果,提高识别准确率,从人类视觉分析的特点出发,首先根据车牌图像中汉字字符和字母数字不同的统计特性对最优的阶梯型边缘检测算法进行了改进;并提出了一种新的动态分割,区域处理的方法。实验表明本方法能够有效提取车牌字符特别是汉字中的较复杂的细节信息,保证了字符边缘的准确性以及内部联通区域的一致性,并对不同质量的车牌图像具有一定的普适性,为识别提供了良好的保证。In order to improve the partition effect of the character letters and the background on license plates in its preprocessing procedure and to enhance the precision rate of the recognition, proceeding from human＇ s vision analysis characteristics, the optimal trapezoidal edge detection algorithm is improved at first in accordance with different statistic traits of the Chinese character and the letter-digitals in vehicle license plate; and then a new method of dynamic segmentation and region process is proposed as well, As experimental results demonstrate, the precision of character edge is ensured, most of available information especially the intricate detail of the Chinese character is preserved and the integrity of character＇ s interior region is maintained by using this new method. Moreover, the new segment method proposed in this article is of universality to a certain extent, which makes a good foundation for license recognition system.

关 键 词：边缘检测 阈值选取 动态区域分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP31[自动化与计算机技术—计算机科学与技术]