Road Segment Interpolation for Incomplete Road Data

Yuya Sasaki, Jiahao Yu, Yoshiharu Ishikawa

Abstract: Road data is fundamental information for location-based services. We trust that the road data is complete to represent an actual road network when we develop the location-based services. However, road data may be incomplete due to update delays, and thus location-based services may not provide useful results. Several algorithms have been proposed to automatically update road data. In this paper, we study interpolation of missing road segments by using vehicle trajectory data. We can find missing road segments from the trajectories because vehicles may pass through road segments that are not included in road data. However, trajectories are inherently noisy due to GPS errors. Hence, we cannot easily interpolate appropriate road segments. We propose an algorithm based on map matching and clustering techniques for achieving accurate and comprehensive interpolation. Our algorithm first detects trajectories that are probably on missing road segments. It then clusters the trajectories by DBSCAN and integrates the trajectories for interpolating the road data. Through the experiments using real incomplete road data and trajectory data, we verify that our algorithm effectively interpolates the missing road segments.