Information extraction and geolocalization of historical aerial imagery

Abstract

Antarctica plays a crucial role in global climate change research, but current models only extend back to the satellite era from the 1990s to the present, leaving significant gaps in our understanding of the continent’s past and covering only limited locations. Historical aerial surveys, which may extend back to the 1930s and cover more locations, offer an opportunity to overcome these limitations. However, creating models from historical imagery poses a significant challenge. This study presents a workflow for automatically extracting camera altitude information from historical imagery and doing preliminary geolocalization by computing the footprints of the imagery based on the extracted altitude information. The workflow was applied to two flight lines in the TMA Single Frames dataset, and the results were evaluated using a scoring system and by comparing the photos with their footprints. This work serves as a preliminary stage in geolocating and georeferencing historical imagery, laying the foundation for future model-building with these photographs.