Issue |
ITM Web Conf.
Volume 70, 2025
2024 2nd International Conference on Data Science, Advanced Algorithm and Intelligent Computing (DAI 2024)
|
|
---|---|---|
Article Number | 01017 | |
Number of page(s) | 6 | |
Section | Traffic Prediction and Analysis | |
DOI | https://doi.org/10.1051/itmconf/20257001017 | |
Published online | 23 January 2025 |
Traffic Sign Recognition in Rainy Conditions Based on Federated Learning
Computer Science, Wuhan University of Science and Technology, 430065 Wuhan, China
Corresponding author: cyl202113407434@wust.edu.cn
The challenge of traffic sign recognition in rainy conditions poses significant difficulties for autonomous driving systems, primarily due to obscured visibility and altered sign characteristics. To tackle this issue, this paper simulated rainy environments to improve the recognition accuracy of traffic signs in real world. This paper utilized OpenCV to preprocess images by adding a rain effect, thereby enhancing the dataset’s realism. Subsequently, this study implemented a LeNet model within a Federated Learning framework, which enables decentralized training while preserving data privacy. The approach involved leveraging the Belgium Traffic Sign Classification Benchmark dataset, achieving an impressive accuracy of approximately 93% in recognizing traffic signs despite the simulated rainy conditions. The federated learning model effectively aggregated knowledge from multiple clients, resulting in a more resilient and efficient recognition system. The proposed method is demonstrated by experimental results to enhance performance in challenging weather conditions while also maintaining data privacy in machine learning applications. Overall, this paper underscores the potential of integrating federated learning with CNNs to improve traffic sign recognition capabilities.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.