Space Archaeology: Survey and Implementation of Deep Learning Methods for Detecting Ancient Structures

Remote sensing instruments are changing the nature of archaeological work. No longer are archaeological discoveries done by field work alone. Light Detection and Ranging, or LiDAR, optical imagery and different types of satellite data are giving opportunities for archaeological discoveries in areas which might be inaccessible to archaeologists. Different types of machine learning and deep learning methods are also being applied to remote sensing data, which enables automatic searches to large scale areas for detection of archaeological remains. In this thesis faster R-CNN object detection deep learning frameworks were used to train models and apply these to three types of archaeological remains. LiDAR based Digital Terrain Models were used to identify burial mounds in Norway. Optical imagery was used to identify fortress structures in Central Asia. Synthetic Aperture Radar data, or SAR, was used to detect archaeological settlement mounds in Central Asia. The success and limitations of these models are presented