车牌图像的快速匹配识别方法  

A Fast Matching Method of Car Plate Recognition

作　　者：胡爱明[1] 周孝宽[1] 

机构地区：[1]北京航空航天大学图像中心,北京100083

出　　处：《计算机工程与应用》2003年第7期90-91,110,共3页Computer Engineering and Applications

摘　　要：在车牌识别中,经常遇到车牌字符发生缺损、污染现象,统计识别、结构识别方法很难将它们准确识别出来。该文提出了一种快速匹配识别方法,根据图像放大和缩小原理构建多种字符模板,利用图像上、下矩对字符进行粗分类,然后根据图像自相关特性,构建了失配比例加权惩罚匹配模型,可以对车牌字符进行精确识别,该方法识别率达到了99%以上。The stained character or fractured character often appear in a collected car license plate image,so structural and statistical character recognition method are not suitable for an applied car license plate recognition system.A template matching based fast car license plate recognition system is presented in this paper.The required template character can be generated from standard character by way of expanding and shrinking.Using a two stage classifier,first order moment for rough classifying and an improved match model-mismatch penalizing model-for precise classifying,a car license plate image recognition system is constructed.The accuracy of recognition can reach99%.

关 键 词：车牌图像 快速匹配识别方法 人工神经网络 图像识别系统 汽车车牌识别系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] U492.22[自动化与计算机技术—计算机科学与技术]