Research on Steel Rail Surface Rolling Defect Detection Using the Fusion of Inductive Thermography and Optical Imaging Technologies
e-Journal of Nondestructive Testing · 2026-07 · 31 (7)
DOI: 10.58286/33121 ↗Kurz gefasst
Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Kang Tian und 4 weiteren Autor:innen.
Abstract (Originalsprache)
To address the challenge of detecting rolling defects on rail surfaces, this paper proposes a spatio-temporal multimodal fusion network, STMM-FuseNet, based on inductive thermography and optical imaging. The network jointly models the infrared temporal temperature field and the spatial texture in infrared/visible images, and incorporates a physics-informed neural network constraint derived from the heat-conduction partial differential equation. This enables pixel-wise defect probability prediction and high-resolution overlay visualization. Experimental results show that, compared with single-frame infrared input, temporal modeling, and single-modal convolutional networks, STMM-FuseNet achieves significantly improved detection rates for weak and small defects, as well as stronger background suppression.
Zitieren
APA
Tian, K., Kou, L., Zhao, B., Gao, X., & Peng, J. P. (2026). Research on Steel Rail Surface Rolling Defect Detection Using the Fusion of Inductive Thermography and Optical Imaging Technologies. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33121
Vancouver
Tian K, Kou L, Zhao B, Gao X, Peng JP. Research on Steel Rail Surface Rolling Defect Detection Using the Fusion of Inductive Thermography and Optical Imaging Technologies. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33121
BibTeX
@article{Tian2026,
author = {Tian, Kang and Kou, Liaoyu and Zhao, Bo and Gao, Xiaorong and Peng, Jian ping},
title = {Research on Steel Rail Surface Rolling Defect Detection Using the Fusion of Inductive Thermography and Optical Imaging Technologies},
journal = {e-Journal of Nondestructive Testing},
year = {2026},
volume = {31},
number = {7},
doi = {10.58286/33121}
}
EUVIMED