Computing the Past – Utilizing Historical Data Sources for Map-Based Retrospective Landscape Research

This paper presents a method for the extraction of high level compound forest cover information from manually produced scanned historical topographic maps. The contained forest cover is a unique data source for GIS-based land cover change modelling. As a case study, this work focuses on the so-called Siegfried Map, the Swiss national topographic map series of the 19 century. The map is of low graphical quality which is related to closely spaced map symbols, overlapping /intersecting hand-drawn graphics and text as well as ill-defined transitions to background through blurring. To conceptually order the extraction process, which is achieved by multi-step recognition, the concept of saliency is introduced. By finding salient features in the image the choice of the adequate next step is facilitated. The recognition steps to be carried out are character recognition by two-stage template matching, line detection by sequential line tracking and continuation, and finally structural analysis for hierarchical classification. The extraction of the forest area based on the recognized structures is carried out by semantic expansion whereby the meanings of objects, which are encountered during expansion, trigger different actions. Index Terms – Historical topographic maps, text/graphics separation in maps, line detection, structural pattern analysis, semantic processing, saliency.