Mapping Historic Alaska Glaciers Using Declassified Spy Satellite Imagery and Computer Vision Based Photogrammetry

Abstract

Recent advances in computer vision tools for photogrammetric 3D modeling have yielded new applications in the re-tasking of archival imagery from a variety of sources with elucidating and sometimes unexpected results. The practice of employing archival aerial imagery to provide multi-temporal spatial data has opened new avenues of research for a variety of fields, including archaeology, climatology, hydrology and glaciology. Building 3D digital surface models from archival aerial imagery that can be compared with more recent digital maps of the same areas-of-interest can help identify changes in land forms that are otherwise difficult to measure because of the speed of processes or proximal events and process that have obscured or masked topographic evidence. This thesis explores an approach to extending the geographic record by re-tasking image data from now declassified military reconnaissance satellites for reconstructing historical topography.