基于多尺度卷积神经网络的交通标志识别  

Traffic Sign Recognition Based on Multi-Scale Convolutional Neural Network

作　　者：薛之昕[1] 郑英豪 肖建[1] 魏玲玲[2] XUE Zhixin;ZHENG Yinghao;XIAO Jian;WEI Lingling(School of Information Engineering,Nanchang University,Nanchang 330029,China;College of Information Engineering,Jiangxi University of Technology,Nanchang 330029,China)

机构地区：[1]南昌大学信息工程学院,南昌330029 [2]江西科技学院信息工程学院,南昌330029

出　　处：《计算机工程》2020年第3期261-266,共6页Computer Engineering

基　　金：江西省科技计划重点项目(20181BBG70031)。

摘　　要：基于多纵卷积神经网络的交通标志识别算法识别率较高,但识别和训练时间较长,实用性较差。为此,构造一种基于多尺度卷积神经网络的道路交通标志识别模型。通过改进单尺度卷积神经网络中特征提取的基网络,将网络不同层级所产生的特征融合为多尺度特征并提供给分类器,以提高低层特征的利用率。在GTSRB数据集上的实验结果表明,该模型准确识别率达到99.25%,与多纵卷积神经网络模型相比,其在保证高精度的同时,识别和训练时间的降幅均超过90%,更适用于真实路况下交通标志的精准检测。The traffic sign recognition algorithm based on multi-column Convolutional Neural Network(CNN)has an ideal recognition rate,but its recognition and training time is longer,so its practicability is poorer.Therefore,a road traffic sign detection model based on multi-scale CNN is constructed.By improving the base network of feature extraction in the single-scale CNN,the features generated by different layers of the network are fused into multi-scale features and provided to the classifier,so as to improve the utilization of the lower features.Experimental results on the GTSRB dataset show that the traffic sign recognition of the model is 99.25%.Compared with the multi-column CNN neural network model,while ensuring high accuracy,the recognition and training time decreases by more than 90%,which is more suitable for the accurate detection of traffic signs under real road conditions.

关 键 词：多层特征 多尺度卷积神经网络 多纵卷积神经网络 交通标志识别 单尺度卷积神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]