Vehicle trajectory-clustering method based on road-network-sensitive features

Existing trajectory-clustering methods do not consider road-network connectivity, road directionality, and real path length while measuring the similarity between different road-network trajectories. This paper proposes a trajectory-clustering method based on road-network-sensitive features, which can solve the problem of similarity metrics among trajectories in the road network, and effectively combine their local and overall similarity features. First, the method performs the primary clustering of trajectories based on the overall vehicle motion trends. Then, the map-matched trajectories are clustered based on the road segment density, connectivity, and corner characteristics. Finally, clustering is then merged based on the multi-area similarity measure. The visualization and experimental results on real road-network trajectories show that the proposed method is more effective and comprehensive than existing methods, and more suitable for urban road planning, public transportation planning, and congested road detection.

Automated summary

This paper introduces a trajectory-clustering method based on road-network-sensitive features, which effectively addresses the issue of similarity metrics among different road-network trajectories. By combining overall vehicle motion trends and road characteristics, the method clusters trajectories more effectively and improves its overall performance.