Deriving Double-Digitized Road Network Geometry from Probe Data

Increasing availability of probe data sources has the potential for enabling automatic map updates, refine the shape of existing map road geometry as well as estimate basic map attributes. The present paper proposes a comprehensive end-to-end unsupervised method based on principal curves for creating bi-directional road geometry from sparse probe data yielding a complete double-digitized road network from raw probe sources without prior map information. The resulting road segments in the road network graph enable conflation with existing map data to identify map changes including basic map attributes such as direction of travel, turn restrictions and traveled speed.