ArticleOpen Access http://dx.doi.org/10.26855/feee.2025.06.006
Optimization of CNN Models for Traffic Sign Recognition
Jian Sun
Iowa State University, Ames, Iowa 50011, USA.
*Corresponding author: Jian Sun
Published: September 26,2025
Abstract
As a critical component of intelligent transportation systems and autonomous driving technologies, traffic sign recognition directly impacts road safety and traffic management efficiency. This study addresses practical challenges in existing CNN-based traffic sign recognition systems, including insufficient accuracy, high computational complexity, and difficulties in small object detection, proposing a comprehensive optimization solution. By introducing lightweight network architectures and channel attention mechanisms, the approach effectively balances model performance with computational efficiency. Incorporating dis-tinctive visual characteristics of traffic signs, targeted data augmentation strategies are designed to significantly enhance the model's adaptability to special scenarios. The adoption of multiscale feature fusion technology improves recognition effectiveness for small-sized targets. Compared to conventional methods, the proposed optimization solution demonstrates notable advantages in recognition accuracy, computational efficiency, and robustness, providing reliable technical support for practical deployment of intelligent transportation systems and holding significant practical value for advancing autonomous driving technologies.
Keywords
Traffic sign recognition; Convolutional neural network; Model lightweighting; At-tention mechanism; Multi-scale features; Intelligent transportation
References
[1] Li X. Research on traffic sign recognition based on deep learning [master's thesis]. Jilin Institute of Chemical Technology; 2024.
[2] Li D. Research on road traffic sign recognition methods and experimental system development [master's thesis]. Tianjin University of Technology and Education; 2022.
[3] Zhai F. Research on traffic sign recognition based on deep learning [master's thesis]. Anhui Jianzhu University; 2020.
[4] Luo F. Pedestrian flow traffic state recognition technology incorporating improved FCM algorithm. Electron Des Eng. 2025;33(17):67-70+75.
[5] Jia Y, Yang Z, Peng G, et al. An optimized algorithm for traffic sign recognition based on LeNet-5 network. Internet Things Technol. 2025;15(15):18-22.
[6] Wang X, Feng F. Research on an intelligent transportation system for the Internet of Things based on RFID technology. Internet Things Technol. 2025; Epub ahead of print.
[7] Yang D. Spatiotemporal pattern recognition of terminal area air traffic flow based on improved DBSCAN clustering. J Civ Aviat Flight Univ China. 2025;36(4):27-31+46.
[8] Li X. Research on the identification of the subject of tort liability for autonomous vehicle traffic accidents. J Lianyungang Tech Coll. 2025;38(2):29-37.
How to cite this paper
Optimization of CNN Models for Traffic Sign Recognition
How to cite this paper: Jian Sun. (2025) Optimization of CNN Models for Traffic Sign Recognition. Frontiers in Electrical and Electronic Engineering, 1(1), 25-29.
DOI: http://dx.doi.org/10.26855/feee.2025.06.006