基于视频的道路交通标志的检测与分割  

Detection and Segmentation of Road Traffic Signs Based on Video

作　　者：赵晓娜 王夏黎[1] 武琦 王博学 ZHAO Xiao-na;WANG Xia-li;WU Qi;WANG Bo-xue(School of Information Engineering,Chang’an University,Xi’an 710064,China)

机构地区：[1]长安大学信息工程学院,陕西西安710064

出　　处：《计算机技术与发展》2018年第12期71-73,78,共4页Computer Technology and Development

基　　金：国家自然科学基金(61473220);中国博士后科学基金(2012M521729)

摘　　要：基于视觉的计算机交通标志识别是现代智能交通系统的重要组成部分之一,其有效运用不仅有利于交通智能化和自动化的管理,更有利于安全性驾驶,是智能交通研究的一个重要内容。其中,交通标志检测是实现交通标志识别的前提,是要解决的关键性问题。目前复杂的道路状况使得交通标志的背景极其复杂,加之车辆行驶过程中的抖动以及光照条件的无形变化等,这些都会导致交通标志失真甚至变形。因此,针对交通标志检测中实时性和准确性的局限,提出一种快速有效的同时采用颜色和形状线索的道路标志检测方法。该方法已经交通标志数据集上进行过测试,收益率平均为94%。该方法对光照不良、部分遮挡、旋转等多种不利情况具有较好的鲁棒性。Vision-based computer traffic sign recognition is an important part of modern intelligent transportation system and its effective use not only benefits traffic intelligent and automatic management,but also is conducive to safe driving,which is an important content of intelligent traffic research.Among them,the traffic sign detection is the premise of traffic sign recognition,which is the key problem to be solved.The current complex road conditions make the background of traffic signs extremely complex,combined with the jitter in the vehicle driving and the invisible changes of light conditions which will cause the traffic signs to be distorted and even distorted.Therefore,in view of the limitation of the real-time and accuracy of traffic sign detection,we propose a rapid and effective road sign detection method using color and shape clues simultaneously.This method has been tested on the traffic sign dataset with an average yield of94%,which is remarkably robust to many adverse conditions such as poor lighting,partial occlusion and rotation.

关 键 词：HSV色彩空间 交通标志检测 形状检测 运动目标检测 智能交通 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]