基于聚类分析的个性化美国车牌分割算法  

Segmentation method for personalized American car plate based on clustering analysis

作　　者：李旭[1] 徐舒畅[2] 尤玉才[3] 张三元[3] 

机构地区：[1]浙江警察学院实验中心,浙江杭州310053 [2]杭州师范大学信息科学与工程学院,浙江杭州310036 [3]浙江大学计算机科学与技术学院,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2012年第12期2155-2159,共5页Journal of Zhejiang University：Engineering Science

基　　金：浙江省自然科学基金资助项目(Y1100557);广东省教育部产学研结合资助项目(2010B090400193;2011B090400546)

摘　　要：针对美国车牌个性化严重,车牌上字符个数、字体、间距和背景等信息都不一致的情况,提出一种可处理复杂多变车牌的车牌分割算法.基于动态的字符分布信息计算车牌倾斜角度和垂直投影局部梯度,利用聚类方法去除非字符区域,并动态确定字符宽度,获得准确字符区域.基于局部梯度的循环分割得到准确的字符分割结果.为了验证该算法,基于1万多张美国车牌的数据集进行实验,结果表明:与已有算法相比,该算法对于具有字符个数不定、字符间距不一致、背景复杂等特征的个性化美国车牌的分割效果有较大提高,分割正确率提高了约5%.A new segmentation algorithm was proposed to handle complex and variant American car license plate, which is very personalized and un-uniform in character number, font, spacing and background. Firstly, the tilt angle and vertical project local gradient were obtained based on the dynamic character distribution information. Then the character width was dynamically calculated and accurate character regions were obtained after applying cluster method to remove non-character regions. Finally, the characters were correctly separated by repeat segmentation based on local gradient. A database with more than 10 000 American plate images was used to verify the accuracy of the proposed algorithm. The experimental results show that compared to existing algorithms, the proposed algorithm can achieve well improved segmentation accuracy up to 5 ~.

关 键 词：个性化车牌 车牌分割 美国车牌 车牌识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]