A Holistic Workflow for Semi-automated Object Extraction from Large-Scale Historical Maps

Abstract

The extraction of objects from large-scale historical maps has been examined in several studies. With the aim to research urban changes over time, semi-automated and transferable holistic approaches remain to be investigated. We apply a combination of object-based image analysis and vectorization methods on three different historical maps. By further matching and georeferencing an appropriate current geodataset, we provide a concept for analyzing and comparing those valuable sources from the past. With minor adjustments, our end-to-end workflow was transferable to other large-scale maps. The findings revealed that the extraction and spatial assignment of objects, such as buildings or roads, enable the comparison of maps from different times and form a basis for further historical analysis. Performing an affine transformation between the datasets, an absolute offset of no more than 72 m was achieved. The outcomes of this paper, therefore, facilitate the daily work of urban researchers or historians. However, it should be emphasized that specific knowledge is required for the presented subjective methodology.