Automatic Defect Detection in Canvas Painting Using Active Infrared Thermography and an Improved YOLOX Network
e-Journal of Nondestructive Testing · 2026-07 · 31 (7)
DOI: 10.58286/33589 ↗At a glance
Journal article in e-Journal of Nondestructive Testing, 2026-07. Written by Zhiyang Zhang and 1 co-authors.
Abstract (original language)
Early-stage defects such as cracks, cuts and internal splitting in canvas paintings are often invisible to the naked eye but may progressively threaten the integrity of artworks. Active infrared thermography (IRT) provides a non-contact means for revealing subsurface anomalies; however, defect interpretation is still largely dependent on expert experience. This work presents a compact framework for automatic defect detection in canvas paintings by combining active IRT with an improved YOLOX object detection network enhanced by a convolutional block attention module (CBAM). To address the scarcity of labeled experimental data in cultural heritage applications, finite-element simulations are used to generate training data, while experimental infrared thermograms acquired from a realistic painting replica are reserved exclusively for testing. Results show that the attention-enhanced YOLOX model significantly improves detection accuracy for typical defects while maintaining a lightweight architecture and near real-time inference speed, demonstrating its potential for in-situ inspection of painted cultural heritage.
Cite
APA
Zhang, Z., & Zhang, H. (2026). Automatic Defect Detection in Canvas Painting Using Active Infrared Thermography and an Improved YOLOX Network. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33589
Vancouver
Zhang Z, Zhang H. Automatic Defect Detection in Canvas Painting Using Active Infrared Thermography and an Improved YOLOX Network. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33589
BibTeX
@article{Zhang2026,
author = {Zhang, Zhiyang and Zhang, Hai},
title = {Automatic Defect Detection in Canvas Painting Using Active Infrared Thermography and an Improved YOLOX Network},
journal = {e-Journal of Nondestructive Testing},
year = {2026},
volume = {31},
number = {7},
doi = {10.58286/33589}
}
EUVIMED