Research on Visibility Level Classification Method of Local Fog Based on Residual Network

DOI: https://doi.org/10.62517/jbdc.202601212

Author(s)

Canwei Weng, Tianyuan Liu

Affiliation(s)

School of Artificial Intelligence and Big Data, Henan University of Technology, Zhengzhou, Henan, China

Abstract

As a meteorological phenomenon with extremely strong locality, suddenness, and small spatial scale, local fog poses severe challenges to highway traffic safety. Traditional meteorological observation methods and satellite remote sensing technologies struggle to capture its dynamic changes in a timely and accurate manner. Current monitoring systems face issues such as high equipment investment, limited monitoring accuracy, and poor warning timeliness. To address these problems, this paper proposes a local fog visibility level classification method based on Residual Network (ResNet). According to meteorological industry standards, local fog is divided into four levels: less than 50 meters (severe local fog), 50 to 200 meters (moderate local fog), 200 to 500 meters (light local fog), and 500 to 1000 meters (slight local fog). By constructing a deep residual network model and utilizing the residual learning mechanism to effectively alleviate the gradient vanishing problem, efficient feature extraction and accurate classification of local fog images are achieved. To comprehensively evaluate model performance, Convolutional Neural Network (CNN) and AlexNet are introduced as comparison models for systematic experiments. Experimental results show that the ResNet model demonstrates optimal comprehensive performance in local fog recognition tasks, achieving a test set accuracy of 95.45%, with particularly excellent discrimination ability for light and moderate local fog. This research provides an effective technical solution for intelligent monitoring and early warning of local fog on highways, and has significant value for improving road traffic safety under adverse weather conditions.

Keywords

Local Fog Monitoring; Deep Learning; Image Classification; Residual Network; Visibility Classification

References

[1] Liang L, Tian H, Li A X, et al. Analysis of Local Fog Weather Characteristics and Discussion of Meteorological Services in Yingshang, Anhui on November 15, 2017. Meteorological and Environmental Sciences, 2020, 43(01): 9-17. [2] National Development and Reform Commission, Ministry of Transport. Implementation Plan for Promoting "Internet Plus" Convenient Transportation. E-Government, 2016, (08): 118. [3] Jiang Z Q, Men A D, Wang H Y. Deep Learning in Computer Vision. Publishing House of Electronics Industry, 2021: 329. [4] Zhou Y S. Research on Application of Image Recognition Technology Based on Deep Learning Algorithm. Beijing University of Posts and Telecommunications, 2017. [5] Bao Y, Shen Q, Xue J, et al. Denoising Diffusion Spatial-Temporal Residual Multi-Graph Convolutional Network for Traffic Flow Prediction. Applied Soft Computing, 2025, 182: 113656-113656. [6] Yu H, Zhang Y L. Research on "Local Fog" Detection Method Based on CNN-SVM. Computer Knowledge and Technology, 2025, 21(08): 1-4. [7] Ni G Y, Tang Z M. Highway Local Fog Detection Method Based on Deep Learning. Computer and Digital Engineering, 2023, 51(12): 2971-2976. [8] Wang J. Research and Application of Sea Fog Detection Model Based on Time Series and Deep Learning. Nanchang University, 2023. [9] Kang J, Miyazaki H, Kim H S, et al. Multi-Satellite Image Matching and Deep Learning Segmentation for Detection of Daytime Sea Fog Using GK2A AMI and GK2B GOCI-II. Remote Sensing, 2025, 18(1): 34-34. [10]Wang F F, Liao B, Du Z J, et al. Research on Classification of Local Fog Early Warning Grade on Guizhou Expressway. Science and Technology Vision, 2024, 14(31): 122-124. [11]HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.