Searching for Viking Age Fortresses with Automatic Landscape Classification and Feature Detection

Abstract

Across the world, cultural heritage is eradicated at an unprecedented rate by development, agriculture, and natural erosion. Remote sensing using airborne and satellite sensors is an essential tool for rapidly investigating human traces over large surfaces of our planet, but even large monumental structures may be visible as only faint indications on the surface. In this paper, we demonstrate the utility of a machine learning approach using airborne laser scanning data to address a "needle-in-a-haystack" problem, which involves the search for remnants of Viking ring fortresses throughout Denmark. First ring detection was applied using the Hough circle transformations and template matching, which detected 202,048 circular features in Denmark. This was reduced to 199 candidate sites by using their geometric properties and the application of machine learning techniques to classify the cultural and topographic context of the features. Two of these near perfectly circular features are convincing candidates for Viking Age fortresses, and two are candidates for either glacial landscape features or simple meteor craters. Ground-truthing revealed the latter sites as ice age features, while the cultural heritage sites Borgø and Traelbanke urge renewed archaeological investigation in the light of our findings. The fact that machine learning identifies compelling new candidate sites for ring fortresses demonstrates the power of the approach. Our automatic approach is applicable worldwide where digital terrain models are available to search for cultural heritage sites, geomorphological features, and meteor impact craters.

Urbanization and Riverine Hinterlands: A Proposal for an Integrative High-Definition and Multi-Scalar Approach to Understanding Ancient Cities and their Dynamic Natural Resources

“…Developing a toolkit for automatic feature detection and pattern recognition is of great importance for archaeo logists globally, as pressure on cultural heritage from climate change, conflict, and development mean we lack the resources to map the ever-accumulating archives of remotely sensed imagery (Bennett, Cowley, and De Laet 2014;Sevara and others 2016). As has recently been shown, this is feasible on national scales in Europe using Random Forests Regression (Stott, Kristiansen, and Sindbaek 2019).…”

“…Recent advances in image processing have made possible the generation of high-resolution 3-D digital terrain models (DTMs) from sequences of archival aerial and satellite images (Risbøl and others 2015;Sevara and others 2018). Using a combination of aerial images and CORONA satellite imagery from the 1960s and 1970s and comparing these to modern SAR-and LiDAR-derived DTMs will provide a comprehensive overview of how topo graphy has changed (Stott, Kristiansen, and Sindbaek 2019;Tapete and Cigna 2019), and provide insight into processes of erosion and deposition. Additionally, the historic DTMs provide topo graphic measurements before the modern intensification of urban development obscured evidence of diagnostic topographic features indicative of large-scale fluvial, colluvial, and tectonic processes.…”

Crossing the Maelstrom: New Departures in Viking Archaeology

“…For Borgring and Gokstad, the critical factor in pointing to unexpected spots for excavation were remote sensing and geophysical surveys, which helped bridge the gap between landscape studies and excavations. Artificial intelligence techniques such as automatic landscape classification and feature detection may further enhance the use of these data (Stott et al 2019). In a similar way, geoarchaeology may refocus attention to hitherto neglected activities (Macphail et al 2013;Milek 2012;Milek and Roberts 2013) and site history (Cannell et al 2016;Devos et al 2013;Macphail and Linderholm 2016;Wouters et al 2016).…”

Employment, Utilization, and Development of Airborne Laser Scanning in Fenno-Scandinavian Archaeology—A Review

“…Alternative visualization methods like local dominance, sky-view factor, and local relief model have seen limited use in archaeology in Denmark, although they offer considerable potential [57,60], (Figure 4a,b). Local dominance in particular proved optimal in the very flat Danish terrain.…”

“…The discovery of the Danish Viking Age ringfort Borgring on the national hillshade model sparked a search for further undetected forts [94]. To reduce the immense task of a systematic manual search, computer vision algorithms were used in a semi-automated search for additional Danish Viking Age ringforts based on the 0.45 point/m 2 DEM from 2007 [60]. These ring detection efforts initially resulted in the identification of 202,048 circular features, later reduced to 199 sites after further processing by using machine learning (among other means) to classify the cultural and topographic context.…”