AI-driven Structure Detection and Information Extraction from Historical Cadastral Maps (Early 19th Century Franciscean Cadastre in the Province of Styria) and Current High-resolution Satellite and Aerial Imagery for Remote Sensing

内容摘要

Cadastres from the 19th century are a complex as well as rich source for historians and archaeologists, whose use presents them with great challenges. For archaeological and historical remote sensing, we have trained several Deep Learning models, CNNs as well as Vision Transformers, to extract large-scale data from this knowledge representation. We present the principle results of our work here and we present a the demonstrator of our browser-based tool that allows researchers and public stakeholders to quickly identify spots that featured buildings in the 19th century Franciscean Cadastre. The tool not only supports scholars and fellow researchers in building a better understanding of the settlement history of the region of Styria, it also helps public administration and fellow citizens to swiftly identify areas of heightened sensibility with regard to the cultural heritage of the region.