一种新的车牌字符快速识别方法

针对车牌字符图像的特点,在研究各种特征提取方法适用场合的基础上提出了改进的特征提取方法和字符识别方法。对于英文和数字,使用一种基于模板匹配和神经网络的车牌字符识别方法。该方法采用两级识别,第一级采用模板匹配识别差别明显的字符,第二级采用BP神经网络识别第一级不能确定的相似字符。对于汉字采用小波变换和LDA提取特征。该方法利用小波变换的特性最大程度地提取了字符图像的特征信息。实验结果表明此算法具有较高的识别率和较快的识别速度。

A Novel Method Of Quickly Recognizing Characters On Vehicle License Plate

This paper proposes a method of character recognition on vehicle license plate based on the research of kinds of feature extraction methods applying situation and characteristic of vehicle license image. The method of recognition of Alphabetic and digit characters is based on pattern matching and NN.This method has two steps.The first step recognizes characters which have distinct difference by using template matching.The second step uses BP neural network to recognize similar characters which can't be recognized in the first step.The wavelet transform and LDA are used for Chinese characters recognition.This method extracts character image's feature information fully depending on wavelet transform.Experimental result shows the algorithm has high recognition rate and fast speed.