Towards the automated large-scale reconstruction of past road networks from historical maps

• DOI: 10.1016/j.compenvurbsys.2022.101794
• Publication Date: 2022-03-18T19:03:36Z
• Authors: Johannes H. Uhl; Stefan Leyk; Yao-Yi Chiang; Craig A. Knoblock
• Keywords: Georeference; Digital mapping; Street network

Transportation infrastructure, such as road or railroad networks, represent a fundamental component of our civilization. For sustainable planning and informed decision making, thorough understanding the long-term evolution transportation infrastructure networks is crucial. However, spatially explicit, multi-temporal network data covering large spatial extents are scarce rarely available prior to 2000s. Herein, we propose framework that employs increasingly scanned georeferenced historical map series reconstruct past by integrating abundant, contemporary color information extracted from maps. Specifically, method uses segments analytical units extracts roads inferring their existence in based on image processing clustering techniques. We tested over 300,000 representing more than 50,000 km United States, extending across three study areas cover 42 topographic sheets dated between 1890 1950. evaluated approach comparison other datasets against manually created reference data, achieving F-1 scores up 0.95, showed statistics highly plausible time, i.e., following general growth patterns. demonstrated geospatial integrated with open new avenues for quantitative analysis urbanization processes landscape changes far beyond era operational remote sensing digital cartography.