Improving Place Name Recognition Accuracy of Historical Topographic Maps

(18-06-2020) Improving place name recognition accuracy of historical topographic maps

Scanned historical topographic maps contain valuable geographical information. Often, these maps are the only reliable source of data for a given period. They are typically only annotated with a title (general location), a date, and a small description. This gives little to no information about what is depicted on the map itself. To extract additional information from these maps, they still need to be analyzed manually, which can be very tedious.

This work uses open-source text recognition tools to recognize coordinate information surrounding the raster maps, in order to automatically georeference the maps. The place names on the maps were recognized and geocoded as well, to provide additional metadata. Because of the large text recognition error rate, additional techniques were developed to improve its accuracy and provide meaningful information on what is depicted on the map. Furthermore, a small study was performed to study the effects of the scan quality and image size on the text recognition accuracy.

 This is research of Kenzo Milleville Steven Verstockt and Nico Van de Weghe , research group IDLab in cooperation with the Department of Geography.



The surrounding map coordinates can be used to georeference the map and improve the place name recognition