微型电脑应用2025，Vol.41Issue(5)：17-21,5.
基于轻量级卷积神经网络的交通标志检测与识别研究
Research on Traffic Sign Detection and Recognition-based on Lightweight Convolutional Neural Network
摘要
Abstract
With rapid development of cities,the problem of road congestion becomes more and more serious,and the issue of traffic sign detection and recognition(TSDR)attracts much attention.In the complex light environment,the traditional TSDR method is not accurate,and it is difficult to give consideration to both recognition accuracy and real-time performance.In view of the above problem,a TSDR system is constructed based on traffic sign detection(TSD)of multi-feature fusion and traffic sign recognition(TSR)of lightweight convolutional neural networks(CNN).The experimental results show that the accuracy,recall and F1 of TSD algorithm based on multi-feature fusion are 98.47％,97.24％and 97.86％,respectively.The accuracy of TSR-Net-FL model based on lightweight CNN is 1.08％higher than that of AlexNet-FL model,and the time is reduced by 9.99 ms.To sum up,the TSDR system based on multi-feature fusion and lightweight CNN has high accuracy and real-time performance.关键词
交通标志检测/轻量级CNN模型/TSDR系统/Retinex算法Key words
traffic sign detection/lightweight CNN model/TSDR system/Retinex algorithm分类
信息技术与安全科学引用本文复制引用
廖惠敏,刘文平,董婉青,刘思琦,董雷..基于轻量级卷积神经网络的交通标志检测与识别研究[J].微型电脑应用,2025,41(5):17-21,5.基金项目
北京市交通委员会科技项目(BJJTW-2022-DSJ-05) （BJJTW-2022-DSJ-05）
