Automatic extraction of former WWI battlefields from ancient maps - Université de Picardie Jules Verne Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Automatic extraction of former WWI battlefields from ancient maps


The former battlefields of World War I (WWI) provide an interesting framework for studying the long-term impacts of ancient anthropogenic disturbances on current ecosystem functioning. The 47 map sheets of devastated regions at 1:50,000 scale edited in 1920 by the geographic service of French army locate the areas heavily damaged by trenches and bombing, the destructed cities, roads and forests inventoried at the end of WWI. As they stand, these scanned maps are not usable under a geographic information system (GIS). A protocol was implemented on 5 sheets to compare the effect of two transformation models (thin plate spline and polynomial order 3) and the number of ground control points on the quality of georeferencing. A second protocol based on morphological operators, color space transformation and K-means clustering classification was tested on 12 different map sheets to extract areas heavily damaged and figures of punctual destructions. Neither significant effects of transformation model or number of ground control points were confirmed. The local thin plate spline method exacerbates non-natural local distortions linked to the research of ground control points and the simplification of physical objects on the maps. With polynomial order 3 transformation and 50 ground control points, residuals vary from 35 to 70 m depending on the map. The second protocol extracted accurately data of interest, with an accuracy varying between 0.31 and 100% depending on data to extract and their presence or absence on the map sheets. The resulting shapefiles are now available and workable in GIS.
Fichier non déposé

Dates et versions

hal-04521338 , version 1 (26-03-2024)



Nelly Paradelle, Marianne Laslier, Guillaume Decocq. Automatic extraction of former WWI battlefields from ancient maps. Remote Sensing for Agriculture, Ecosystems, and Hydrology XXV, Sep 2023, Amsterdam, Netherlands. pp.18, ⟨10.1117/12.2684009⟩. ⟨hal-04521338⟩
8 Consultations
0 Téléchargements



Gmail Facebook X LinkedIn More