Research on Pavement Crack Monitoring Technology in Urban Road Based on Environment Perception‐Driven Visible Light‐Infrared Fusion
Structural Control and Health Monitoring · 2026-01 · 2026 (1)
DOI: 10.1155/stc/3339613 ↗Kurz gefasst
Fachartikel in der Zeitschrift Structural Control and Health Monitoring, online erschienen am 28.06.2026. Verfasst von Hui Wu und 6 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Department of Transportation of Hunan Province.
Abstract (Originalsprache)
Urban road crack hazard monitoring is a critical part of ensuring sustainable development of cities. However, existing UAV inspection methods mostly rely on visible light single‐modal images, which are highly susceptible to interference from complex environments such as illumination changes, rain, fog, and snow, severely limiting their robustness. This research proposes an environment perception‐driven end‐to‐end visible light‐infrared (RGB‐IR) fusion segmentation architecture (ED‐DualFuseSeg) model. The core innovation lies in a hybrid fusion strategy driven by environment perception, which quantifies environmental interference and image quality, dynamically allocates dual‐modal weights at the feature layer, and designs a continuously differentiable dynamic threshold generation mechanism at the decision layer, forming a bidirectional linkage anti‐interference mechanism between feature fusion and decision optimization. Comprehensive experiments on the PIRDSD, DeepCrack, and GAPs384 datasets demonstrate that RGB‐IR dual‐modal fusion outperforms single modality, and the proposed ED‐DualFuseSeg model significantly improves upon existing SOTA models. Specifically, on PIRDSD, it achieves a precision of 83.15% and an mIoU of 79.90%; on DeepCrack, an mIoU of 88.04%; and on the high‐noise GAPs384 dataset, it attains an excellent balance with an mIoU of 70.29% and a recall of 68.14%, exhibiting strong environmental adaptability and robustness. Ablation experiments further verify that the collaborative design of the two core modules significantly reduces the false positive rate, and the dynamic weight allocation strategy markedly enhances the model’s adaptability to complex environments. The research indicates that this method effectively improves the hazard monitoring precision and robustness of pavement cracks in urban roads, providing technical support for high‐frequency and wide‐coverage inspections by UAVs. Dataset is available at https://doi.org/10.5281/zenodo.19568642 .
Förderung
- Department of Transportation of Hunan Province
Zitieren
APA
Wu, H., Hu, K., Li, S., Duan, J., Lu, Y., Zheng, W., & Xie, J. (2026). Research on Pavement Crack Monitoring Technology in Urban Road Based on Environment Perception‐Driven Visible Light‐Infrared Fusion. Structural Control and Health Monitoring, 2026(1). https://doi.org/10.1155/stc/3339613
Vancouver
Wu H, Hu K, Li S, Duan J, Lu Y, Zheng W, et al. Research on Pavement Crack Monitoring Technology in Urban Road Based on Environment Perception‐Driven Visible Light‐Infrared Fusion. Structural Control and Health Monitoring. 2026;2026(1). doi:10.1155/stc/3339613
BibTeX
@article{Wu2026,
author = {Wu, Hui and Hu, Kaiyuan and Li, Shaopeng and Duan, Jin and Lu, Yi and Zheng, Wei and Xie, Jianbin},
title = {Research on Pavement Crack Monitoring Technology in Urban Road Based on Environment Perception‐Driven Visible Light‐Infrared Fusion},
journal = {Structural Control and Health Monitoring},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/stc/3339613}
}
EUVIMED