作者：黄文杰

作者单位：淮阴工学院交通工程系, 江苏淮安 223003

关键词：归一化; 特征提取; 聚类分析; 神经网络; 字符识别

摘要：

在车牌字符识别的实际应用中,只采用改进的BP神经网络算法不能彻底改变其固有的算法缺陷。因此,重点阐述采用聚类分析与神经网络的方法分别对车牌中汉字和英文字母及阿拉伯数字进行识别,以加快车牌的识别速度,以适应高速公路收费系统即时、准确的要求。实验结果表明,对数字与字母的识别率为97.0%,对汉字的识别率为90.1%,识别时间小于3s,既兼顾了BP神经网络识别的稳定性,又考虑到高速公路收费的实时性需要。

Method and application of license plate recognition base on cluster analysis

HUANG Wen-jie

Department of Transportation Engineering, Huiyin Institute of Technology, Huai'an 223003, China

Abstract: In the application of license plate recognition, the inherent defaults of the improved BP neural network algorithm couldn't be overcome thoroughly when it was only used. Thus, the Chinese characters and the English letters plus the numbers should be recognized separately with the methods of the cluster analysis and neural network independently, which would accelerate the speed of the license plate recognition in order to satisfy the demands of immediacy and accuracy in the system of expressway toll-collection. The experimental results showed that the recognition accuracies of letters and numbers were 97.0%, that of Chinese characters was 90.1%, and the time for recognizing each license plate was less than 3 s. This method utilized the stability of the recognition algorithm of BP Neural Network, and took the real-time need of expressway toll into consideration at the same time.

Keywords: Normalization; Extracting characteristic; Cluster analysis; Neural network; Character recognition

2008, 34(4): 76-80  收稿日期: 2007-11-27;收到修改稿日期: 2008-2-4

基金项目: 

作者简介: 黄文杰(1977-),男,江苏南京市人,硕士,专业方向为图像识别、图像重构、模式识别、交通工程。

参考文献

[1] Zou X T, Chen X Q, Chen J. An intelligent vehicle license plate recognition system[J], Proceedings of SPIE-The International Society for Optical Engineering, 2001, 4556(3):15-19.
[2] Hu Y F, Zhu F, Zhang X D. A novel approach for license plate recognition using subspace projection and probabilistic neural network[J], Lecture Notes in Computer Science, 2005, 3497(5):216-221.
[3] Juntanasub R, Sureerattanan N. Car license plate recognition through hausdorff distance technique[C]. 17th IEEE International Conference on Tools with Artificial Intelligence, 2006, 6(1):5-8.
[4] ShapiroV,GluhchevG.Multinational license plate recognition system[C]. Proceedings-International Conference on Pattern Recognition, 2004, 4(3):352-355.
[5] Trier O, Jain A K. Taxt torfinn feature extraction methods for character recognition-asurvey[J]. Pattern Recognition, 1996, 29(4):641-662.
[6] 飞思科技产品研究发中心. 神经网络理论与Matlab7实现[M].北京:电子工业出版社, 2005.
[7] 黄文杰, 王海涛, 姬建岗.基于公路收费系统的车牌定位与字符分割算法[J].电视技术, 2006(2):85-87.
[8] 韩笑, 马驷良, 张禹, 等.基于神经网络的机动车号牌字符识别[J]. 吉林大学学报:理学版, 2005, 43(4):461-466.
[9] 陈蕾, 黄贤武, 仲兴荣, 等.基于改进BP算法的数字字符识别[J].微电子学与计算机, 2004, 21(12):127-130.
[10] 邢博, 梁德群, 李文举.一种新的车牌数字及字母字符识别方法[J].辽宁师范大学学报:自然科学版, 2005, 28(1):56-58.
[11] 甘龙, 高隽, 梁栋, 等.基于分形维数的车牌字符识别[J].中国公路学报, 2002, 15(4):75-77.