基于特征投影和交叉覆盖神经网络的车牌识别  

License Plate Location Based on Projection Character and Alternate Covering Neural Network

作　　者：宓浩[1] 张燕平[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《计算机技术与发展》2007年第10期76-79,共4页Computer Technology and Development

基　　金：安徽省自然科学研究基金(050420208)

摘　　要：汽车牌照的自动识别在智能交通系统中占有重要地位,应用前景广阔。在自动识别过程中,牌照中的数字和汉字具有数量少和字体特征固定的特点,故其投影特征明显,利用此性质可以对车牌汉字进行快速分类,但精度不高。神经网络分类准确,且有很强的鲁棒性,但运算量大,识别时间太长且数据不易收敛。文中提出的基于投影和交叉覆盖神经网络的车牌识别方法充分融合利用了两者的优点,克服了各自的不足,达到了较好的结果。The recognition of vehicle license plate plays a very important role in intelligence transportation system. It has wide application ranges. In automatic recognition process, the Chinese character in the car license has the quantity to be few and the font characteristics are inherent, therefore its projection characteristic is obvious. Using projection characteristic to recognize license plate can be used for rapid classification. But the precision is not high. Neural network classification is accurate, and has strong robusmess, but the operation is large, also not easy to restrain. In this paper, license plate location based on projection character and alternate covering neural network method can be used to combine the advantages of them, and overcome the shortcomings of them, so it can achieve better results.

关 键 词：特征投影 汉字识别 车牌识别 交叉覆盖 神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]