摘要
将支持向量机(SVM)引入到复杂条件下运动车辆牌照字符的识别中。回顾了车牌识别研究的现状,简要介绍了SVM的基本原理,比较了SVM算法和神经网络算法在车牌字符识别上的优劣;提出了采用基于先验知识的二叉树结构组合多个二值分类支持向量机来解决车牌字符的多类识别问题。在实验中采用了LibSVM训练软件,针对车牌汉字的小字符集进行了仿真,同时与神经网络分类方法进行了比较。实验结果表明该方法的汉字识别率较高,在小字符集车牌汉字识别中具有较强的实用性。
The application of SVM is presented in vehicle licence recognition. Firstly, the basic of current studies for licence recognition and SVM theories are introduced, and SVM algorithm with nerve network (N-N) algorithm is compared. Then bintrees that based on previous knowledge and multiple classifiers is adopted to resolve recognition for vehicle licence. Software LibSVM is proposed for recognition of vehicle licence in the experiment. The results are also compared with NN classifier, which indicates that the SVM strategy im- prove recognition rate and therefore has more practicability.
出处
《计算机工程与设计》
CSCD
北大核心
2006年第21期4033-4035,4042,共4页
Computer Engineering and Design
基金
河南省杰出人才创新基金项目(0221000200)