摘要
研究车牌字符识别问题,针对传统神经网络在车牌字符识别存在识别准确率低、效率低的问题,提出了一种基于改进神经网络的车牌字符识别方法。该方法首先采用Gabor滤波器提取车牌字符的特征,PCA降维处理消除车牌字符特征之间的冗余信息,然后采用改进的神经网络对提取特征进行训练得到最优识别模型,最后利用最优模型对车牌字符进行识别。仿真实验表明,数字及字母的识别准确率达95.0%以上,汉字的识别准确率达93.1%,与传统识别方法相比,识别准确率和识别速度都有了较大的改进,该方法在车牌识别的应用有着广泛的前景。
n the automatic recognition system of vehicle license,feature selection for vehicle character recognition is the key factor in pattern recognition. In view of the deficiencies of traditional combination optimization method and the shortcoming of too early convergence of simple genetic algorithm,a new method of license plate recognition is proposed. First,the features of plate characters were detected by Gabor filter and the principal component analysis feature extraction. Then the features were used to train the back propagation neural networkclassifier. Finally,the plate characters were classified by the BPNN. Using the algorithm,a high recognition rate can be reached. Experimental results showed that the recognition accuracy of number and alphabet is above 95.0% and that of Chinese character is 93.1%.The algorithm is feasible,robust and applicable.
出处
《计算机仿真》
CSCD
北大核心
2010年第9期299-301,350,共4页
Computer Simulation
关键词
神经网络
车牌字符识别
特征提取
Neural network
License plate recognition
Feature extraction
