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

Machine learning analysis of airborne laser scanning data has identified two promising Viking Age ring fortress sites in Denmark, along with two glacial landscape features initially mistaken for meteor craters. The study employed Hough circle transformations and template matching to detect over 200,000 circular features, which were then narrowed down using geometric properties and machine learning classification. Ground-truthing confirmed Borgø and Trælbanke as viable cultural heritage sites, warranting further archaeological investigation. This approach demonstrates significant potential for global applications in detecting cultural heritage, geomorphological features, and meteor impact craters where digital terrain models are available.

Stott, D., Kristiansen, S. M., & Sindbæk, S. M. (2019). Searching for Viking Age Fortresses with Automatic Landscape Classification and Feature Detection. Remote Sensing, 11(16), 1881. https://doi.org/10.3390/rs11161881