Assessment and Visualization of OSM Building Footprint Quality
ISBN 978-85-88783-11-9
Authors
1Müller, F.; 2Iosifescu Enescu, I.; 3Hurni, L.
1ETH ZURICH Email: fabian.mueller@ethz.ch
2ETH ZURICH Email: iosifescu@ethz.ch
3ETH ZURICH Email: lhurni@ethz.ch
Abstract
OpenStreetMap (OSM) is a crowd sourcing project, mostly known for its open data on street topology and land use. However, since 2007, OSM contributors have started to map also building footprints. In several cities or even countries, these footprint datasets are now deemed to be near completion and increasingly used as an alternative to proprietary data. In this paper we are investigating the completeness and other quality characteristics of the OSM building footprints for an entire country, namely Switzerland, by comparing them with official data sources. The largest provider of proprietary geospatial data in Switzerland is swisstopo, which offers a topographic landscape model (TLM) as a set of vector features with an accuracy of between 0.2 and 1.5 metres for building footprints. The overall quality of OSM data has been the subject of some research especially regarding the street network. However, an automated process to deliver comprehensive quality metrics has rarely been discussed for building footprints and even more scarcely implemented for general use. Moreover, in order to use OSM data and thus profit from their permissive licencing in both science and commerce, a detailed analysis in such a scale that it is applicable for investigation areas of any scope, be it county or even country level, is irremissible, even though, due to the constantly changing nature of OSM data, such analyses do not hold their validity for very long. Using the TLM data as reference, the OSM building footprint data for Switzerland may thus be evaluated in terms of data quality and suitability. To achieve this, a comparison algorithm based on the building footprints’ respective centroid distance and a shape signature function has been developed and implemented in Java using the Open Source Java GIS toolkit GeoTools, reading the adjusted data from a PostGIS database. The specific shape signature function used is the polygons’ turning function Θ, a translation-, scale- and rotation-independent representation of a polygon based on the relative length of each perimeter section and its angle relative to the following perimeter section. The integral of two turning functions may then be used as a polygon distance function and thus—in combination with a preceding selection of potential identical buildings via nearest-neighbour analyses performed on their centroids—as a similarity metric, yielding intuitive results in most cases and allowing building footprints beneath a certain threshold to be defined as identical. The analysis results have been visualised in QGIS with an innovative approach via hexagonal grids incorporating three levels of detail to display areas of higher population density with better accuracy whilst still retrieving interpretable results for areas of lower population density. Both the developed program and the visualisation methodology may be easily adapted for countries for which a reference model of similar quality is available, or to compare different temporal versions of building footprint data. In the spirit of openness promoted by the ICA-OSGeo Open Geospatial laboratory at ETH Zurich, the program will be released to the general public under an open source licence.
Keywords
OSM; building footprint quality; turning function