摘要
汽车牌照的定位是智能交通系统中的重要组成部分之一,其定位效果直接关系到后期的识别工作,应用前景广阔。为了有效实现车牌的准确定位,文中首先在灰度图像中基于提取部分怀疑区域,然后使用基于构造性学习的交叉覆盖算法,对区域样本进行学习后构造出对应的神经网络,然后使用该网络对新进样本进行定位,从怀疑区域中确定出牌照的位置。对不同背景和光照条件下的大量实验结果表明定位准确率较高,从而该方法可行有效,有较强的实用价值。
Location of vehicle license plate is an important component of ITS (intelligent transportation system),because the result of this work has a direct relationship with the later recognition work and has a promising future.In order to locate the plate effectively,this article first extracts the dubious regions from gray scale image and then constructs a neural network via the study of these regions using structural alternative covering algorithm.Then use this neural network to determine the position of new plates. Multiple experiments under different background and lightening conditions prove the result of location is accurate.Thus this method is feasible, efficient and valuable in practical application.
出处
《微机发展》
2004年第8期41-43,46,共4页
Microcomputer Development
基金
国家自然科学基金科研基金资助项目(60175018
60135010)
安徽省教育厅自然科学研究基金资助项目(2002kj112)
关键词
构造性机器学习
交叉覆盖算法
汽车牌照
定位
structural machine learning
alternative covering algorithm
car plate
location
