一种低图像质量车辆牌照的字符分割方法  

A CHARACTER SEGMENTATION METHOD FOR DEGRADED VEHICLE LICENSE PLATE

作　　者：王琰滨[1] 蒋龙泉[1] 冯瑞[1] 

机构地区：[1]复旦大学计算机科学技术学院媒体计算研究所,上海201203

出　　处：《计算机应用与软件》2013年第3期108-110,117,共4页Computer Applications and Software

基　　金：上海市科委科技创新行动计划项目(10511500703);上海市教委科研创新项目(产学研类)(11CXY01);宝山区科委产学研合作项目(CXY-2010-35)

摘　　要：提出一种应用于车牌识别系统中车牌字符分割的方法,能够对由光照不均匀,车牌表面污损等原因造成的低图像质量的车牌进行有效字符分割。首先通过连通区域分析和车牌校正对车牌进行预处理,然后利用车牌固有的先验知识与间隙分析方法,实现整个车牌字符的精确分割,克服污损、边框和倾斜等干扰因素。通过在两组数据集上进行的实验结果表明,该算法具有很强的实用性与鲁棒性。This paper proposes a novel character segmentation method for degraded license plate which are usually caused by uneven light,dirty surface or something else.Firstly,connected component analysis with OCR and plate correction will be used to obtain high confidence character regions.Then according to license plate format prior knowledge,an algorithm based on character gap analysis calculates the positions of rest characters and finally complete the whole segmentation process.This method could effectively avoid the stain,tilt and plate frame＇s bad effect.The experimental results on 2 different data set prove that it is a practical and robust segmentation approach.

关 键 词：低质量车牌 字符分割 连通区域分析 字符识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]