EUVIMEDSkopri EUVIMED
Crossref

Deep learning dased multiclass surface damage detection for RC Bridge inspection using UAV imagery

Ali Siddique, Vittorio Prodromo, Antonio Bilotta

e-Journal of Nondestructive Testing · 2026-07 · 31 (7)

DOI: 10.58286/33591 ↗

At a glance

Journal article in e-Journal of Nondestructive Testing, 2026-07. Written by Ali Siddique and 2 co-authors.

AstrattDettalji biblijografiċiCiteRelated studies

Abstract (original language)

Reinforced concrete (RC) bridges are vital infrastructure assets whose safety and durability depend on regular inspections. Traditional visual inspection methods are time consuming, costly, and prone to subjectivity, highlighting the need for automated solutions. This study proposes a UAV-based multiclass damage detection framework for RC bridges using deep learning and high-resolution drone imagery. To better represent the complex geometry of real-world damage, a polygon-based annotation strategy with pixel-level masks was adopted instead of conventional rectangular bounding boxes. A detailed damage taxonomy consisting of ten classes such as efflorescence, exposed rebar, concrete runoff, mechanical spall, spall, splitting crack, concrete water marks, crack map, surface defect, and mild crack was defined to capture diverse deterioration mechanisms. YOLOv11 detection model were trained and evaluated on the annotated dataset of 600 high resolution images, demonstrating superior performance compared to earlier bounding box based approaches. The model achieved an overall mAP@0.5 of 0.47 across all classes, with a maximum F1-score of 0.51. Precision recall and F1-confidence analyses indicate robust detection performance, particularly for spalling and exposed reinforcement, while maintaining balanced precision and recall across multiple damage categories. The results confirm that fine-grained polygon annotations significantly enhance detection accuracy for small, irregular, and spatially distributed damage.

Cite

APA

Siddique, A., Prodromo, V., & Bilotta, A. (2026). Deep learning dased multiclass surface damage detection for RC Bridge inspection using UAV imagery. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33591

Vancouver

Siddique A, Prodromo V, Bilotta A. Deep learning dased multiclass surface damage detection for RC Bridge inspection using UAV imagery. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33591

BibTeX
@article{Siddique2026,
  author = {Siddique, Ali and Prodromo, Vittorio and Bilotta, Antonio},
  title = {Deep learning dased multiclass surface damage detection for RC Bridge inspection using UAV imagery},
  journal = {e-Journal of Nondestructive Testing},
  year = {2026},
  volume = {31},
  number = {7},
  doi = {10.58286/33591}
}
Start for free →