Abstract
Historic aerial images are invaluable sources of aid to archaeological research. Often collected with large-format photogrammetric quality cameras, these images are potential archives of multidimensional data that can be used to recover information about historic landscapes that have been lost to modern development. However, a lack of camera information for many historic images coupled with physical degradation of their media has often made it difficult to compute geometrically rigorous 3D content from such imagery. While advances in photogrammetry and computer vision over the last two decades have made possible the extraction of accurate and detailed 3D topographical data from high-quality digital images emanating from uncalibrated or unknown cameras, the target source material for these algorithms is normally digital content and thus not negatively affected by the passage of time. In this paper, we present refinements to a computer vision-based workflow for the extraction of 3D data from historic aerial imagery, using readily available software, specific image preprocessing techniques and in-field measurement observations to mitigate some shortcomings of archival imagery and improve extraction of historical digital elevation models (hDEMs) for use in landscape archaeological research. We apply the developed method to a series of historic image sets and modern topographic data covering a period of over 70 years in western Sicily (Italy) and evaluate the outcome. The resulting series of hDEMs form a temporal data stack which is compared with modern high-resolution terrain data using a geomorphic change detection approach, providing a quantification of landscape change through time in extent and depth, and the impact of this change on archaeological resources.