基于最小二乘法和Radon变换的车牌校正算法  

Algorithm of vehicle license plate correction based on least square method and Radon transform

作　　者：陈航[1] 李晓东[1] 李云红[1] 田冀达 

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《西安工程大学学报》2015年第4期404-409,共6页Journal of Xi’an Polytechnic University

基　　金：西安市科技局技术转移促进工程项目(CXY1431(4));陕西省教育厅科研计划资助项目(14JK1319);陕西省科技厅自然科学基础研究计划资助项目(2013JC2-15);国家级大学生项目(201310709004);西安工程大学大学生创新创业训练计划项目(201403018)

摘　　要：为了准确定位车牌识别系统中车牌字符区域,提出基于最小二乘法和Radon变换的车牌校正方法.首先,水平校正采用最小二乘法,先由水平差分法求出车牌字符信息的水平特征点,再通过水平特征点拟合字符的水平倾斜方向,从而求得倾斜角来进行水平校正;然后,垂直方向采用Radon变换,即通过以1°为步长在[-30°,30°]内遍历每个角度,对每个角度进行Radon变换,变换结果一阶导数绝对值累加和最大值对应的角度即为剪切角.对一个包含200幅测试图像的图像库进行测试,有效校正率可达到96%以上,且校正的速度有了一定的提高.实验结果证明,这样既保证了算法的广泛实用性,又改进了算法的实时性.To locate the license plate character region accurately ,an algorithm least square meth‐od and Radon transform for license plate correction was proposed .Firstly ,the license plate was corrected by using the least square method in horizontal direction .The horizontal tilt angle was got by using feature points to fit characters tilt direction .Secondly ,the horizontal corrected li‐cense plate was corrected by using Radon transform in vertical direction .The license plate im‐age was transformed by Radon transform in every angle in the range of [-30＆#176; ,30＆#176;] ,and then the vertical shear angle was corresponding to the angle of maximum value of sum of the abso‐lute value of the first-order derivative of transform results .An image set which contains 200＆amp;nbsp;pictures w as tested ,w hich show s that the correction achieves more than 96% ,and its operating efficiency is improved .Results suggest that this method is not only widely applicable ,but also real time .

关 键 词：车牌校正 HOUGH变换 RADON变换 主成分分析 最小二乘法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]