一种快速精确的汽车牌照字符切分算法  

A Fast Algorithm for Accurate Char Segmentation of License Plate

作　　者：任明亮[1] 范勇[1] 蒋欣荣[1] 游志胜[1] 

机构地区：[1]四川大学图象图形研究所,成都610064

出　　处：《四川大学学报（自然科学版）》2002年第3期450-454,共5页Journal of Sichuan University(Natural Science Edition)

摘　　要：针对汽车牌照具有字符间距、字符数、字体等统一的纹理特征 ,提出了一种快速精确的牌照字符切分方法 ,设计了一系列的特征函数和判决函数突出其纹理特点从而进行车牌字符的切分 .对在不同气候条件和不同地点采集到的 10 0 0幅图片进行了实验 ,实验结果表明 ,该算法具有快速和鲁棒性的特征 ,字符切分准确度达到 98.1% .Char segmentation of license plate rapidly and precisely is the base and precondition of character recognition,and therefore is vital for the automatic recognition of license plate. Characters in License plate have similar texture features. Authors propose a fast algorithm for accurate char segmentation of license plate, and propose feature functions and judge functions to enhance image texture feature. An experiment is carried out with 1000 images which were captured under varied weather conditions and places. The results of experiment show that the algorithm is fast and robust, and accurate segmentation rate is 98.1%.

关 键 词：车牌字符切分 纹理特征 二值化 投影法 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]