Development of an automated matching algorithm to assess the quality of the OpenStreetMap road network : a case study in Göteborg, Sweden

Sammanfattning: In the last decade a new and alternative source of geospatial data has become available, so called Volunteered Geographic Information (VGI). Private individuals voluntarily collect large amount of geographic data for a joint project. This data are usually free under certain licence restrictions. One of the most well known examples of VGI is the OpenStreetMap (OSM) project. Due to its increasing popularity it has developed in less than 10 years to a useful data source. Private, scientific and commercial users are wondering how good is these voluntarily collected data and can it be an alternative to expensive authority and commercial datasets. This study presents a quality assessment of the OSM road network against a reference dataset from Lantmäteriet (Swedish National Mapping Agency). In order to allow a meaningful evaluation, corresponding features in both datasets have to be identified. This process is commonly referred as matching. Difficulties of automatic matching occur due to the different geometric representations of the two datasets. An automated matching algorithm is developed to match road network data from OpenStreetMap and Lantmäteriet. The method is adopted from Koukoletsos et al. (2012) and developed further to incorporate feature correspondence. The matching is based mainly on geometric and attribute (road name) constraints. Besides minor pre-processing steps the algorithm works completely automated and can be applied to any region with data coverage in Sweden. The matching is performed in a case study covering the area of Göteborg, the second largest city Sweden. The created feature correspondence is then used to calculate the quality elements completeness, positional accuracy and thematic accuracy. The matching algorithm returned good results with an acceptable matching error. The quality assessment revealed that OSM can be seen as a proper data source which has some reservation regarding the attribute accuracy.