基于神经网络的车牌识别技术研究

南京邮电大学

硕士学位论文

基于神经网络的车牌识别技术研究

姓名：刘志军

申请学位级别：硕士

专业：电路与系统

指导教师：王厚大

2011-01

摘要

车牌识别技术是现代智能交通系统中最为关键的技术之一，广泛应用于出入口车辆管理、

不停车自动收费、交通流量检测、失窃车辆查询等领域。与传统的车辆管理方法相比，极大地提高了车辆管理的效率与水平，并且节省了大量的人力、物力，实现了车辆管理的规范化、科学化，越来越受到人们的关注，有着广泛的应用前景。

车牌识别系统一般由车牌定位、字符分割和字符识别三部分组成。本文是基于图像处理

和人工神经网络相关理论，对车牌识别过程中的车牌定位、车牌图像预处理、字符分割、特征提取和字符识别等环节进行研究和分析，并在MATLAB环境下进行了仿真实验。

首先，在车牌定位过程中，采用基于颜色特征和灰度跳变特征的算法实现车牌定位。利

用蓝底白色车牌中蓝色的色度饱和度较大的特点，实现了车牌的粗定位。然后利用车牌图像的灰度跳变特征实现车牌的精确定位。

其次，在图像预处理过程中，利用灰度拉伸技术增强图像对比度，然后采用Ostu方法实

现图像二值化。

再次，在车牌字符分割过程中，采用基于灰度跳变和垂直投影法去除车牌边框，然后采

用基于投影结合先验知识的方法实现车牌图像的分割。

最后利用改进后的BP神经网络，针对汉字、字母、字母和数字、数字四种不同的识别

问题，设计了四种不同的分类器对其进行识别。实验结果表明，本文提出的算法定位准确，识别率较高。车牌识别技术是一个涉及图像处理与模式识别等多学科的综合课题，具有重要的理论和现实意义。

关键词：车牌识别；车牌定位；字符分割；字符识别；人工神经网络

ABSTRACT

In the modern Intelligent Transportation System(ITS), License Plate Recognition(LPR) is one

of the most critical technologies and it can be broadly applied in car management at gate, electronic toll collection system, traffic flow control, stolen vehicle inquiry and so on. Compared with traditional vehicles management methods, it greatly improves managing efficiency and level, and saves manpower and material resources, realizes vehicle management standardized and scientific, has attracted more and more attention by people, and has a very broad application.

Generally LPR system consists of three parts, those are license plate location, character

segmentation and character recognition. Based on image procession technology and artificial neural network technology, the paper researches and analyses the license plate location, the image pretreatment, character segmentation, feature extraction and character recognition process of the car license plate character recognition and has carried on simulation in the environment of MATLAB.

Firstly, in the location process, adopts a license plate technology based on color character and

the frequency of gray change. The paper has realized the thick location of the license plate by using the S value of the blue color is bigger in the blue bottom and white word license plate. Then, according to the character of license plate image’s frequency of gray change, the paper has realize the accurate location of the license plate.

Secondly, in the part of image preprocess, the paper has enhanced image contrast by stretching

the gray of image, used Ostu method to change greysacle image into binary image.

Thirdly, in the part of character segmentation, based on the gray change property and vertical projection threshold, the paper has used a method to eliminate the borders of plate image. Segments characters of license plate by vertical projection and combining with the priori knowledge of plate.

Finally, aiming at four kinds of different recognition problems of Chinese character, the letter,

letter of figure，figure, the paper has designed four kinds of different classification devices to recognize them by using the improved BP neural network. Experiment results show that the method this paper presents has some advantage such as accurate location, high recognition rate. LPR is a comprehensive research item, which involves digital image processing and pattern recognition, and is far-reaching significance in theory and application.

Keywords: license plate recognition; license plate location; character segmentation; character

recognition; artificial neural network

南京邮电大学学位论文原创性声明

本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得

的研究成果。尽我所知，除了文中特别加以标注和致谢的地方外，论文中不包

含其他人已经发表或撰写过的研究成果，也不包含为获得南京邮电大学或其它

教育机构的学位或证书而使用过的材料。与我一同工作的同志对本研究所做的

任何贡献均已在论文中作了明确的说明并表示了谢意。

研究生签名：_刘志军_ 日期：_2011年3月15日__

南京邮电大学学位论文使用授权声明

南京邮电大学、中国科学技术信息研究所、国家图书馆有权保留本人所送

交学位论文的复印件和电子文档，可以采用影印、缩印或其它复制手段保存论

文。本文电子文档的内容和纸质论文的内容相一致。除在保密期内的保密论文

外，允许论文被查阅和借阅，可 …… 此处隐藏：11023字，全部文档内容请下载后查看。喜欢就下载吧 ……