基于组合支持向量机的车牌字符识别  

Combination of support vector machine for character recognition of license plate

作　　者：施隆照[1] 强书连 

机构地区：[1]福州大学物理与信息工程学院,福建福州350108

出　　处：《计算机工程与设计》2017年第6期1619-1623,共5页Computer Engineering and Design

摘　　要：为克服传统支持向量机(SVM)算法应用在车牌字符识别问题上出现的结构复杂、训练样本庞大、识别速度慢的问题,提出一种多分类支持向量机结构,即组合多分类支持向量机(CMM-SVM),应用于车牌字符识别。根据车牌字符分布特点,设计3组基于超球面的一类支持向量机(OC-SVM),分别识别汉字字符、字母字符、数字与字母混合字符,依据距离最小原则进行初步识别;在前一步识别的基础上,采用标准SVM对易混淆的字符进一步判断,提高识别率。通过大量的样本测试验证了该结构能够准确识别复杂环境下的各类字符,识别率高,识别速度快。To overcome the problems such as complex structure, huge amount of training samples and more time-consumption ap-peared in the traditional support vector machine （SVM） used for character recognition of license plate, an multi-classification SVM structure , namely combination of multi-class support vector machine （CMM-SVM） was proposcharacter recognition of license plate. According to the features of the license plate , three types of one-class SVM based on hy-per-sphere were designed to recognize Chinese characters , alphabetic characters and comprehensive characters , and each one of them was recognized depending on the minimum distance. Several standard SVMs were established based recognize some characters that could by easily confused to improve the rate. According to the test with a large number of sam-ples ,the proposed structure can recognize the characters accurately in complex environment , and has high recognition rate and recognition speed.

关 键 词：车牌字符识别 一类支持向量机 支持向量机 超球面 组合多分类支持向量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]