摘要
文章研究了车牌识别系统中的字符分割和识别技术。提出一种投影法粗分割结合先验知识后处理的字符分割方法,该方法简单、容易实现,取得了很好的分割效果。对于字符识别,本文采用SVM(SupportVectorMa鄄chine)方法,并根据车牌字符特征将子分类器分为四组,提高了识别率、缩短了训练时间,实验表明,用该方法识别车牌字符具有较高的识别率和识别速度,并避免了神经网络局部极值等问题。
This paper mainly researches on character segmentation and recognition techniques which are used in license plate recognition system. A character segmentation method which uses projection to segment primarily then combine with apriori knowledge to dispose behind is proposed. This method is simple, easy to implement, and have a good effect. For character recognition, this paper applies SVM (Support Vector Machine) method, Sub-classifiers are divided into four groups to improve recognition accuracy and shorten training time. Experimental results demonstrate that using this method to recognise license plate characters can get a high recognition accuracy and speed, moreover, it avoids the local extremum problem existed in neural network.
出处
《微电子学与计算机》
CSCD
北大核心
2005年第6期34-36,共3页
Microelectronics & Computer
关键词
字符分割
SVM
字符识别
BP神经网络
Character segmentation, SVM(Support Vector Machine), Character recognition, BP neural network