基于深度学习的交通标志识别算法  

Traffic sign recognition algorithm with deep learning

作　　者：徐仙伟[1] 曹霁[1] Xu Xianwei;Cao Ji(Information Technology Department,Nanjing Forest Police College,Nanjing,Jiangsu 210023,China)

机构地区：[1]南京森林警察学院信息技术学院

出　　处：《计算机时代》2019年第6期67-70,共4页Computer Era

基　　金：江苏高校品牌专业建设工程资助项目(PPZY2015A058);中央高校基本科研项目(LGYB201507)

摘　　要：智能车在移动时需要对周围环境进行快速有效的认知,文章在规则环境下对视觉中所包含的信息理解标志牌等进行研究,提出了一种基于深度学习的交通标志识别算法。该算法先进行边缘检测,提取出图像中交通标志所在的兴趣区域;再利用LeNet-5深度学习算法,对已提取出来的兴趣区域进行识别。以德国标准交通标志图数据库为实验数据的结果表明,该方法能有效理解交通标志信息,具有识别效率高、使用简单等特点。Smart cars need fast and effective recognition of the surrounding environment when moving.This paper studies the information understanding signs contained in the vision of rule environment,and proposes a traffic sign recognition algorithm with deep learning.Firstly,edges in image are detected to get the interesting area which includes the traffic sign;then,a trained LeNet-5 deep learning algorithm is used to recognize it.The experiment results on benchmark images of German traffic sign database show that this method can effectively understand traffic sign information,and has the characteristics of high recognition efficiency and easy to use.

关 键 词：交通标志 环境理解 边缘检测 深度学习 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]