Automatic Update of Road Attributes by Mining GPS Tracks

Despite advances in cartography, mapping is still a costly process which involves a substantial amount of manual work. This article presents a method for automatically deriving road attributes by analyzing and mining movement trajectories (e.g. GPS tracks). We have investigated the automatic extraction of eight road attributes: directionality, speed limit, number of lanes, access, average speed, congestion, importance, and geometric offset; and we have developed a supervised classification method (decision tree) to infer them. The extraction of most of these attributes has not been investigated previously. We have implemented our method in a software prototype and we automatically update the OpenStreetMap (OSM) dataset of the Netherlands, increasing its level of completeness. The validation of the classification shows variable levels of accuracy, e.g. whether a road is a one-or a two-way road is classified with an accuracy of 99%, and the accuracy for the speed limit is 69%. When taking into account speed limits that are one step away (e.g. 60 km/h instead of the classified 50 km/h) the classification increases to 95%, which might be acceptable in some use-cases. We mitigate this with a hierarchical code list of attributes.