Three-dimensional structure determination of grade-separated road intersections from crowdsourced trajectories

Abstract

Although existing research on road intersection detection has been widely conducted using sensor data, mapping grade-separated road intersections in three-dimensions is still lacking. In this study, we propose a novel strategy to obtain detailed three-dimensional structures on grade-separated road intersections at turning levels using crowdsourced trajectories. Using the preprocessed trajectories, we identified grade-separated road intersections and obtained their boundary information based on the designed ECPC (Elevation Changing Points Clustering) algorithm. The three-dimensional structures of these identified grade-separated road intersections, e.g., centerline and vertical order of road segments at turning level, and positions of on-ramp and off-ramp, were extracted through trajectory data incremental integration and elevation trend analysis. We applied the tracking data collected by crowds in Beijing and Shanghai to estimate the effectiveness of the proposed method. The experimental results showed that the average precision, recall, and F1-score of grade-separated road intersection identification in the cities of Beijing and Shanghai were (83.31%, 80.67%, 81.97%) and (75%, 71.43%, and 73.17%), respectively. The accuracy of road centerline, slopes, vertical order of road segments, and positions of on-ramp and off-ramp in Beijing and Shanghai were (95%, 95.9%, 96.7%, 89.6%) and (94%, 87.3%, 87.1%, 87.1%), respectively.

Keywords

• Grade-separated road intersection
• Road network generation
• Urban transport
• Crowdsourced mapping