一种基于特征的BP神经网络车牌字符识别  

License Plate Character Recognition Based on BP Neural Network

作　　者：马晓娟[1] 周钢[1] 潘仁龙[1] 

机构地区：[1]贵州民族学院,贵州贵阳550025

出　　处：《佳木斯大学学报（自然科学版）》2009年第6期831-833,共3页Journal of Jiamusi University：Natural Science Edition

基　　金：校级自然科学类重点资助科研项目(2008XS28)

摘　　要：字符识别是自动车牌识别系统中很关键的一步.字符识别有以下几步,首先,对车牌图像进行预处理.其次,通过竖直方向投影分割字符.最后,将提取的字符特征输入网络进行训练.在实验中,利用该方法对光照不均、字符大小不一、运动背景的图像,特别是相似字符的识别获得了较高的识别率,并且将其与字符输入BP神经网络进行对比分析.实验结果表明,该方法对字符识别有很好的鲁棒性、有效性.Character recognition plays an important role in the automatic license plate recognition（ALPR） system.The license plate recognition was done in the several steps.First,the license plate was preprocessed by image processing algorithms.Then,each character was separated by the vertical projection algorithms.In the last step,all of characters set of license plate were recognized and the features were extracted from each character to lump into a vector as input of the BP neural network.The system worked under variable illumination,variable size of plate and dynamic backgrounds,especially similar characters.The correct recognition rate was very high.Besides,BP neural network train sample attained from the character itself was compared and analyzed.The experimental results demonstrated the great robustness and efficiency of the method.

关 键 词：车牌识别 字符识别 特征提取 BP神经网络 

分 类 号：U491.116[交通运输工程—交通运输规划与管理]