改进特征的车牌识别算法  

License plate recognition algorithm based on improved feature

作　　者：蒋玲玲[1] 杨冬[1] 张友能[2] 

机构地区：[1]淮南联合大学机电系,安徽淮南232001 [2]安徽工贸职业技术学院电气与信息工程系,安徽淮南232001

出　　处：《安庆师范大学学报（自然科学版）》2017年第2期63-65,75,共4页Journal of Anqing Normal University(Natural Science Edition)

基　　金：安徽省高校自然科学研究项目(KJ2015A367;KJ2017A582)

摘　　要：由于光照不均、倾斜、模糊、字符笔画粗细不均匀、切分位置偏差因素,现有车牌识别算法的最终字符识别正确率较低。对现有几种BP字符识别算法所选取的输入特征进行改进和融合,作为BP神经网络的输入,以提高识别的准确度。通过对大量样本仿真实验,证明新特征很好地保留了字符的纹理信息,提高了BP网络对畸异字符的适应性,同时提高了综合识别率,有较高的实用价值。Due to the uncertainties of uneven illumination, tilt, fuzzy, uneven character contrast and location deviations, the indication of the current license plate recognition algorithms become low accuracy. In this paper, based on the comparison of the existing license plate recognition algorithms based on the BP neural network, we improve and mix the improved features, and make them as our proposed new algorithm＇s import, so as to improve the accuracy. Through a set of simulations, we find that the proposed algorithms maintain the texture information of the characters, and the BP network＇s adaptability is improved. In addition, the comprehensive recognition of the characters has been improved, and the proposed algorithms have higher practical values.

关 键 词：车牌识别 字符识别 BP神经网络 改进特征 

分 类 号：TP317[自动化与计算机技术—计算机软件与理论]