基于边缘特征和神经网络的汽车牌照定位算法  

A Vehicle License Plate Location Algorithm Based on Image Edge Features and Neural Network

作　　者：邱刚[1] 王养利[1] 

机构地区：[1]西安电子科技大学,陕西西安710071

出　　处：《微机发展》2005年第4期30-32,共3页Microcomputer Development

摘　　要：汽车牌照定位是一个较难解决的图像分割问题,神经网络为此问题的解决提供了一个有力工具。文中提出了一种新的基于字符边缘特征的定位算法,它通过滑动窗口抽取样本并输入神经网络,对比其输出的特征向量来描述图像中以滑窗左上角顶点为标记的不同位置的边缘特性,结合统计优选的方法提取车牌。实验结果表明:该车牌定位算法识别精度高,速度快。Automatic location of vehicle license plates is a difficult image segmentation problem.Neural network is a powerful tool for solving the problem.In this paper,an algorithm based on neural network and image edge features is presented.It gets out the samples through a slide window and puts them into the neural network.The output eigenvectors of a BP net is used to describe the characteristics of edges at various places where can be marked with the top left corner of the slide window in an image.Combined with statistics,this algorithm can detect possible license plates in an image very precisely and efficiently.

关 键 词：神经网络 图像处理 BP神经网 汽车牌照定位 模式识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]