车牌汉字识别方法研究.doc

车牌汉字识别方法研究
THE RESEARCH OF LICENSE PLATE CHARACTER RECOGNITION
专业:计算机科学与技术
姓名:王锐强
指导教师:
申请学位级别:学士
论文提交日期:2011年6月10日
学位授予单位:天津科技大学
摘要
目前字符识别技术已经相对成熟,字符包括了字母、数字和汉字,由于汉字自身的复杂结构,汉字识别的难度相对比较大,尽管如此还是有很多成熟的算法能够保证在正常图像质量条件下可以取得很高的正确识别率。但对车牌识别来说,在实际系统中获得的车牌的大小、位置、方向不确定,而且由于光照原因、车牌有污点等原因往往会造成汉字的变形、不连续等问题,从而往往难以识别,识别结果难以让人满意。车牌汉字识别方法研究就是研究能够在汉字变形、不连续的情况下能准确的识别出汉字。
本文针对已分割完成的车牌汉字,提出了一种基于正交Gegenbauer 矩特征提取方法,并将其应用于车牌汉字识别中。Gegenbauer 矩这种方法有别于其他传统的汉字特征提取方法,它能获得汉字的全局特征信息,因此可以唯一地表征一个汉字。为了提高车牌汉字识别率,将支持向量机SVM方法用于车牌汉字的识别,算法首先采用Gegenbauer 矩的方法提取车牌汉字图像的特征参数,然后采用提取的特征训练SVM分类器,再应用SVM分类器分类和判别车牌汉字。
关键词:车牌汉字识别; 特征值提取; Gegenbauer矩; 支持向量机
ABSTRACT
The technology of character recognition is now relatively refined, the characters include letters, numbers and Chinese characters, and it is relatively difficult to recognize a Chinese character due to plex structure. There are still many mature algorithms that can guarantee high recognition rate in the presence of the images of normal quality. But for license plate recognition, because of the uncertain of the obtained size, location, direction of the license plate in the actual system, and because of the problem of the deformation and discontinuity of character caused by light, stain on license plates and so on, it is difficult to recognize or get the satisfactory results. The research of license plate character recognition is to study how we can accurately identify the Chinese characters in the existing of the deformation and discontinuity.
Specific to the divided license plate characters, this article proposes a extraction method based on the feature of the Gegenbauer moment and applies it to the license plates recognition. Gegenbauer moment is different from other traditional feature extraction method; a character can be uniquely characterized as it can obtain the information on global features of Chinese characters. To improve the rate of license plate character recognition, support vector machine is used for license plate charact