EUVIMEDObjevte EUVIMED
Open access · CC BY 4.0

UAV Hyperspectral Corrosion and Coating Characterization of Post-Mining Steel Infrastructure Using Hierarchical Spectral Libraries

Hernan Flores, Bodo Bernsdorf, Tobias Rudolph, Jörg Benndorf

Mining · 2026-11-14 · 6 (4) · 107

DOI: 10.3390/mining6040107 ↗

At a glance

Journal article in Mining, published online on 2026-11-14. Written by Hernan Flores and 3 co-authors. 24 references. The full text is freely available (open access, CC BY 4.0). Funded by: RAG-Stiftung.

AbstraktBibliografické údajeCiteRelated studies

Abstract (original language)

Corrosion and coating degradation on post-mining steel infrastructure are commonly assessed through close-range visual inspection, which can be limited by access constraints, observer subjectivity, and the heterogeneous character of weathered steel surfaces. This study develops a UAV hyperspectral workflow for corrosion and coating characterization on steel infrastructure, combining hierarchical spectral libraries with Spectral Angle Mapper (SAM) classification. Data were acquired with a HAIP BlackBird V2 sensor mounted on a DJI Matrice 350 RTK platform using both nadir and oblique viewing geometries, and were processed into reflectance-calibrated hyperspectral cubes through a standardized preprocessing workflow. A grouped global spectral library derived from six reference steel samples provided a conceptual anchor, while a site-specific library for the Itzenplitz headframe was constructed through a staged, criteria-driven procedure: manual delineation of the visible structure, unsupervised clustering within the delineated area, consolidation of candidate clusters using a fixed spectral-angle threshold, acceptance of classes based on within-class angular dispersion, cross-scene verification between two training scenes, and targeted region-of-interest refinement for spatially localized surface states. The resulting library comprises seven surface-state classes with a minimum inter-class spectral angle of 0.097 rad and within-class dispersion between 0.048 and 0.090 rad. Applied to four field scenes, the library assigned 80.6% (STRONG), 95.4% (PAINTED), 66.0% (EARLY), and 84.3% (COMPLEX) of the analyzed pixels to a class (assigned coverage, which is not a classification accuracy), with the two scenes used for library construction showing higher assigned coverage than the two held-out scenes. Classification was independently corroborated by a two-band Redness Index, which reproduced the paint–corrosion separation without reference to the SAM library in both training and held-out scenes. The classes correspond to interpretable degradation mechanisms described in the corrosion literature, including uniform field corrosion, run-off staining, and blistered corrosion crust, and additionally resolve a secondary yellow-ochre coating system applied to auxiliary walkway elements. The study provides a reproducible framework for spectral-library-based corrosion interpretation on post-mining infrastructure and identifies the validation steps required for operational monitoring.

Funding

  • RAG-Stiftung

Cite

APA

Flores, H., Bernsdorf, B., Rudolph, T., & Benndorf, J. (2026). UAV Hyperspectral Corrosion and Coating Characterization of Post-Mining Steel Infrastructure Using Hierarchical Spectral Libraries. Mining, 6(4), 107. https://doi.org/10.3390/mining6040107

Vancouver

Flores H, Bernsdorf B, Rudolph T, Benndorf J. UAV Hyperspectral Corrosion and Coating Characterization of Post-Mining Steel Infrastructure Using Hierarchical Spectral Libraries. Mining. 2026;6(4):107. doi:10.3390/mining6040107

BibTeX
@article{Flores2026,
  author = {Flores, Hernan and Bernsdorf, Bodo and Rudolph, Tobias and Benndorf, Jörg},
  title = {UAV Hyperspectral Corrosion and Coating Characterization of Post-Mining Steel Infrastructure Using Hierarchical Spectral Libraries},
  journal = {Mining},
  year = {2026},
  volume = {6},
  number = {4},
  pages = {107},
  doi = {10.3390/mining6040107}
}
Start for free →