APP下载

基于特征匹配算法的交通标志牌检测与识别

2018-07-28秦松陈小玉韩鹏贾小林

电脑知识与技术 2018年11期

秦松 陈小玉 韩鹏 贾小林

摘要:随着社会经济发展,机动车数量增加,智能交通系统将成为时代发展的必然要求。而交通标志识别是智能交通系统的重要环节,更是车辆自动导航的前提,具有重要的研究意义和应用价值。针对4类交通标志,即禁止掉头、禁止前行、禁止左转、禁止右转,从以下几个方面进行研究:(1)基于RGB颜色空间进行阙值分割;(2)通过标志连通域比例进行提取;(3)基于现实情况,构造多场景下的训练集,进行训练识别。实验结果显示,该算法可有效的识别出不同光照条件下发生变形、缩放及旋转后的交通标志牌,交通标志牌识别准确率达到了92%。

关键词:交通标识; 标识提取; 标识分割; SVM

中图分类号:TP302 文献标识码:A 文章编号:1009-3044(2018)11-0192-04

Traffic Signboard Detection and Recognition Based on Feature Matching Algorithm

QIN Song, CHEN Xiao-yu ,HAN Peng ,JIA Xia-Lin*

(College of Computer Science and Technology,SouthWest University of Science and Technology,Mianyang 621000,China)

Abstract:With the development of society and economy, the increase in the number of motor vehicles, the intelligent transportation system will become an inevitable requirement for the development of the times. Traffic sign recognition is an important part of ITS, and it is also a precondition in vehicle automatic navigation. So it has important research significance and application value. This article aims at four types of traffic signs,namely, prohibiting U-turn, prohibiting forward turn, no left turn, no right turn, and the following aspects are studied: (1) Threshold segmentation based on RGB color space (2)Extraction based on the proportion of connectivity domain(3)Based on the actual situation, the training set under multi-scene is constructed and trained.The experimental results show that the algorithm can effectively identify the traffic signs with deformation, scaling and rotation under different light conditions,and t the traffic sign recognition accuracy reaches 92%.

Key words: traffic identification; dentification extraction; identification segmentation; SVM

1引言

智能交通系統(Intelligent Traffic System,简称ITS),是将先进的科学技术有效地集成运用于整个交通运输管理体系,而建立起来的一种在大范围内发挥作用的,高效的综合运输管理系统。在智能交通系统中,如何识别交通标识是非常重要的。交通标识的识别要求分类正确,速度快,对图像的平移、旋转变换,光照强度,天气因素不敏感。

近几年来,基于视频交通标志牌的研究有了很大进展,各种算法被相继提出。文献[1]提出了一种通过特征互补方式来提高标志图像的识别方法,但是在干扰情况下,准确率会降低。文献[2]采用模块匹配的方法可以识别出不同类型的交通标志,但是由于模板数量之多,在识别时不具有实时性。文献[3]基于不变矩和SVM方法对原形标志图像进行识别,具有良好的分类能力,具有较好的识别率,但仅对圆形交通交通标志有效,不具有实用性。……

登录APP查看全文