摘要
为解决普通支持向量机多类分类器对车牌字符识别准确率低、速度慢等问题,研究了基于支持向量机二叉分类树的车牌字符识别算法。根据车牌字符的结构特征提出了利于字符分类的粗像素特征提取方案,并对字符进行相应的特征提取,通过KL变换对生成的特征向量进行降维处理以提高字符识别速度,最后利用Fisher判别准则构造支持向量机二叉分类树,保证每类字符均具有最大可分离性,提高了字符识别率。对车牌字符集进行了识别测试,实验结果表明了该算法的可行性和有效性。
To solve the problems of low accuracy and speed of License Plate character recognition based on general support vector machine multi-class classifier effectively,an algorithm of character recognition based on support vector machine binary classification is studied.Firstly,the scheme of coarse pixel feature extraction is proposed,which is helpful for character classification according to the structure of character,and then,character image features are extracted by coarse pixel feature algorithm and the primitive feature vectors will be acquired and decreased by KL transformation to improve the speed of character recognition.Finally,the Support Vector Machine binary classification trees based on Fisher criterion are built,which can guarantee the maximal class separability of every character set and improve recognition accuracy.In the experiment of character recognition,the results show the feasibility and effectiveness of the algorithm.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第9期3166-3169,共4页
Computer Engineering and Design
关键词
支持向量机
特征向量
字符识别
KL变换
二叉树
support vector machine
feature vector
character recognition
KL transformation
binary tree
