一种新的车牌数字及字母字符识别方法  

A New Method for Number and Letter Character Recognition of License Plate

作　　者：邢博[1] 梁德群[1] 李文举[2] 

机构地区：[1]大连海事大学信息工程学院,辽宁大连116026 [2]辽宁师范大学计算机与信息技术学院,辽宁大连116029

出　　处：《辽宁师范大学学报（自然科学版）》2005年第1期56-58,共3页Journal of Liaoning Normal University:Natural Science Edition

基　　金：国家自然科学基金资助项目(60272017)

摘　　要：车牌字符识别是车牌自动识别系统的关键环节.车牌数字及字母字符识别是车牌字符识别的重要组成部分.传统的模板匹配法的缺点是在车牌图像质量退化时识别率较低.提出一种新的改进的模板匹配法.该方法定义前景匹配度、背景匹配度和整体失配度3种匹配测度进行模板匹配,匹配成功则直接获得识别结果,否则,对相似字符情形应用特征区域模板匹配法,对其他情形应用空心模板匹配法.实验结果表明,该方法对车牌字符褪色、污迹、断裂、模糊、光照不均等质量退化车牌图像具有较好的识别效果.Character recognition is a critical step in the automatic license plate recognition system. It is the important part of the character recognition that the recognition of numbers and letters in the license plate. The low recognition rate to the poor vehicle license plate images is weakness of the traditional template match method. In this paper, a new improved pattern match method is presented Three kinds of match measure, the foreground match measure, the background match measure and the integer mismatch degree, are defined to carry on template match. If the match succeeds , the recognition result will be attained directly, otherwise the characteristic district pattern match will be applied to the similar characters and the hollow template pattern match will be applied to other conditions. Experimental results show that the proposed method in this paper has better recognizing effect when dealing with depigmentizing, dirt, split, blurry and uniform-illuminated Number and Letter characters.

关 键 词：模板匹配法 车牌图像 匹配度 车牌字符识别 识别率 车牌自动识别系统 污迹 字母 数字 方法 

分 类 号：N945[自然科学总论—系统科学] TP391[自动化与计算机技术—计算机应用技术]