Multisensor multitemporal AI approaches to the detection and analysis of archaeological settlements at a continental scale
Place: Large Lecture Room
Abstract: The last few years have seen an important increase of AI-based methods for the detection and analysis of non-readily visible archaeological sites and features. However, many of these methods’ results tend to overfit the training data and are largely dependent upon specific cultural and environmental circumstances. This paper will present current efforts to develop a semi-automated workflow for the large-scale detection of archaeological sites within a range of environments. These efforts have focussed on: 1) the development of continental site detection algorithm; and 2) an automated drone-based survey method. The continental site detection algorithm uses the location of known sites to train a machine learning probabilistic classifier, which is able to adapt to specific local environmental conditions. The classifier uses multi-temporal and multi-sensor satellite data in combination with specific raster products that map physical and environmental features of archaeological sites to boost their detection. Despite the potential of automated site detection methods, many ancient settlements have not left any detectable trace beyond the distribution of archaeological materials (mostly pottery fragments) over the surface of the site. In order to address this, an automated drone-based intensive survey method has been developed that is able to map and extract as vectors a high percentage of all visible pottery fragments distributed across the surface of a given study area. This method makes use of photogrammetric processes, machine learning algorithms and geospatial analyses. It was designed for use with standard commercial drones, though the implementation of new drone technologies is able to boost detection rates and significantly reduce false positives. This presentation will discuss theoretical and practical approaches to the satellite and drone-based detection of archaeological sites and explore problems and potential solutions for the development of a global, self-evaluating algorithm implemented in such a way that can be employed by local communities, authorities and development agencies to manage non-visible archaeological heritage.
Short Bio: I obtained my PhD on high mountain archaeology at the Catalan Institute of Classical Archaeology (ICAC). Later on I held postdoctoral positions at the GEOLAB (CNRS, UMR 6042) and the universities of Nottingham, Sheffield and Cambridge. I am now a Ramón y Cajal researcher at ICAC where I co-direct the Landscape Archaeology Research Group. My research currently focus on the development of computational approaches to archaeological research using remote sensing in combination with geospatial techniques, network analysis and artificial intelligence.