基于K-L变换和LS-SVM的车牌字符识别新方法

A new License Plate Character Recognition method Based On K-L transform and LS-SVM

针对车牌识别中识别率和识别速度难以同时提高这一难点,本文提出一种K-L变换和最小二乘支持向量机相结合的车牌字符识别新方法。首先使用K-L变换对预处理后的车牌字符图像进行特征降维;然后根据车牌字符的排列特征采用聚类分析中类距离思想,设计四组最佳二叉树的最小二乘支持向量机子分类器来分别实现字母、数字和汉字的识别。实验结果表明,该方法所设计的分类器较好的解决了传统多类算法中存在的不可分区域情况,具有较高的识别率和识别速度及分类推广能力。

This paper applies constructing and combining several binary least square support vector machines （LS-SVM） with a binary tree, preferably solve the problem that improving the recognizing rate and speed at the same time, firstly the K？L transform is used to low- er the dimension of the features of the image which is after the pretreatment. Sub-classifiers are divided into four groups to identify the letters, numbers and Chinese characters, which is based on binary tree using class distance of clustering . Experimental results show that, the proposed method resolve the unclassifiable regions that exist in the conventional muhiclass svms, get better recognition rate and speed, and can solve the multi？classification issue.