Spatial Accuracy Assessment of Buildings in Openstreetmap

Abstract

The aim of this paper is to assess the spatial accuracy of OpenStreetMap (OSM) with respect to the Turkey Topographic Vector Database (TOPOVT) within the context of ‘building’ layer. Being an open-platform, anyone can access to OSM and add geographic entities as well as update them. Since there is no stringent standards, spatial accuracy assessment of OSM is an open research area. TOPOVT, on the other hand, is produced by the General Directorate of Mapping by following a standard procedure, where the maps are produced for 1:25000 scale or larger scale. Updating this database is a costly process and could only be conducted at specific time intervals. Therefore, automatic detection of the locations requiring update in TOPOVT would be an effective operation, which would eventually reduce the overall cost of the database update. However, the spatial accuracy of the geographical features have to be analysed in order to support such a motivation. Therefore, the aim of this paper is to assess the spatial accuracy of ‘building’ layer by calculating the Hausdorff distance between the matching (homologous) polygons in OSM and TOPOVT. The proposed methodology consists of two methods to detect the matching polygons: ‘overlap method’ and ‘centroid method’. Hausdorff distance is calculated for only those intersecting buildings in both of the layers. Since it is safe to assume that the intersecting polygons refer to the same geographic object, the calculated distance could be used to indicate the spatial accuracy of the building. The developed software is tested on an urban and a rural environment in Ankara, Turkey. The results indicate that the quality of OSM could well match with TOPOVT. Specifically, the average Hausdorff distance is approximately the same for both of the methods: approximately 9.5 metres. Considering that OSM and TOPOVT are generated through completely different processes’, the spatial accuracy is considered to be ‘good’ and ‘useful’ for many practical and operational purposes. In order to increase the effectiveness of the developed methodology in a real-life context, the whole process is integrated into an ArcMap extension and the code is made available on GitHub.