Orthorectification of large datasets of multi-scale archival aerial imagery: a case study from Türkiye

Abstract

Recent research has unveiled the immense potential of orthorectifying small datasets of scanned historical aerial imagery for multi-temporal environmental and social science research on limited areas such as single landforms, sites, and cities. This article takes this approach further and presents methods that aim for accurate yet efficient orthorectification of large datasets of archival aerial imagery for landscape and regional-scale research. The study draws from a colossal archive in Turkiye, working with over 850 scanned historical aerial photographs at photographic scales of 1:60,000 and 1:30,000 to showcase a workflow that combines structure-from-motion (SfM) and auto-registration techniques to rectify and mosaic the images. This endeavor covers an area spanning 3,600 km(2), and produces a 1:60,000 scale orthomosaic and a 1:30,000 rectified mosaic with spatial resolutions of 1.4 m and 0.8 m, respectively. The Root Mean Square Error (RMSE) values reflect sensitivity to accuracy-efficiency trade-offs of the approach. The results thus demonstrate the potential of these combined methods for producing large coverage orthomosaics for regional-scale analyses, while underscoring the challenges that future research must address.