Automatic detection of archaeological features from airborne laser scanning data using deep learning: Development and evaluation of the ADAF tool
Year: 2026
Type: article
Abstract: The need to use machine learning (ML) in archaeology is constantly increasing due tothe growing availability of large amounts of high-quality airborne laser scanning data(ALS, lidar). The Automatic Detection of Archaeological Features (ADAF) tool wasdeveloped to provide an easily accessible deep learning (DL) software for automatingthe detection of archaeological features from ALS data. The software has a userinterface that requires minimal interaction and no prior knowledge of ML techniques,greatly improving accessibility for the archaeological community.The underlying ML models (MLMs) have been trained on an extensive archive of ALSdatasets in Ireland, which have been labelled by experts with three types ofarchaeological features, namely enclosures, ringforts, and barrows. The corecomponents of the tool are the Relief Visualisation Toolbox (RVT) for processing theALS data into ML-friendly visualisations and the Artificial Intelligence Toolbox for EarthObservation (AiTLAS), which provides access to the DL models, both of which areactively used in the field of aerial archaeology. Although developed for detectingselected archaeological monuments in Ireland, ADAF supports retraining with datafrom other regions and other monument types, enabling its use in archaeologicalcontexts worldwide.The practical application of the software was tested on an active infrastructuredevelopment corridor in Ireland. The results showed a promising 84% recall rate forknown archaeological sites and outperformed human detection by discovering 116potential new archaeological sites that hadn't been noticed during manual inspection.
Language: English
Subfield: Space and Planetary Science
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,910
OpenAlex ID: W7144337187