Annotation-guided multi-scale UAV spectral mapping of a ruined medieval hermitage

Masonry and façade material discrimination using multispectral data

Bibliographic Data

ID23311580
AuthorsGuillem Montalban-Faet (0009-0004-2530-4008, Universitat de València), Rafael Fayos-Jordan (0000-0002-2787-1856, Universitat de València), Enrique A Navarro (0000-0001-8365-0233, Universitat de València), Jaume Segura-Garcia (0000-0002-9138-5465, Universitat de València), Teresa Ferrer-Valls (Universitat de València)
Year2026
Volume17
Issue35
Publication date2026-07-24
Open AccessYes
TypeARTICLE
VenueVirtual Archaeology Review (JOURNAL)
Journal identifiersISSN: 1989-9947 • E-ISSN: 1989-9947
PublisherUniversitat Politecnica de Valencia (PUBLISHER)
DOI10.4995/var.2024.26323
OpenAlexW7171728492
LanguageEN
References cited46

Unmanned Aerial Vehicle (UAV) photogrammetry and multispectral imaging are increasingly used in cultural heritage documentation, but the interpretation of spectral products remains challenging when no independent radiometric calibration or exhaustive field validation is available. This paper presents an annotation-guided, multi-scale UAV workflow for the spectral mapping and characterisation of the ruined Sant Miquel Hermitage in Corbera, Valencia, Spain. Red-Green-Blue (RGB) and four-band multispectral data acquired with a DJI Mavic 3M and processed in DJI Terra are used as scene-relative contrast layers. The workflow is applied at two complementary scales. At the landscape scale, manual annotations of visible masonry are used to generate reference masks for index ranking. The analysis therefore evaluates the ability to discriminate known visible fabric rather than the independent detection of previously unknown archaeological features. At the façade scale, seven annotated material and construction zones are analysed using RGB chromatic indices, multispectral indices, and Grey-Level Co-occurrence Matrix (GLCM) texture features. The results show that composite indices combining brightness, vegetation absence, and microrelief provide the most robust discrimination of visible masonry. Among single spectral indices, vegetation-based contrasts remain highly effective, but new non-vegetation-based formulations based on visible/NIR balance provide an interpretable alternative linked to the spectral behaviour of exposed mineral surfaces. At the façade scale, RGB-derived brightness, greenness, redness, and colouration indices are more informative than multispectral vegetation indices for material discrimination, while GLCM texture adds an independent axis for distinguishing dressed stone, rammed earth, and weathered surfaces. The study demonstrates that commercial UAV datasets can support reproducible heritage interpretation when combined with explicit annotation, comparative index ranking, and careful treatment of uncertainty. Rather than claiming a detection capability equivalent to excavation, the workflow provides a transparent framework for mapping visible masonry, characterising façade materials, and identifying surface anomalies that merit targeted field verification.

Hyperspectral imaging · Multispectral image · Orthophoto · Photogrammetry · RGB color model · Spectroradiometer · Vegetation (pathology) · Workflow · 3D Surveying and Cultural Heritage · Archaeological Research and Protection · Remote Sensing and LiDAR Applications

Tools

Open DOIOpen Access