Vehicle trajectories clustering plays an increasingly essential role in understanding urban traffic patterns. In order to improve the clustering effect, in this paper, we propose a novel clustering method integrated with multidimensional trajectory information based on ST-OPTICS clustering algorithm. This density-based algorithm utilizes spatial, temporal, road segment and direction angle information to form clusters of varying density based on spatial and temporal closeness. Furthermore, we utilize DTW to build a similarity model to measure the similarity between vehicle trajectories. Then we conduct experiments on a large-scale vehicle trajectory dataset consisting of 2172 trajectories collected from the GPS traces nearby Beijing Olympic Parks. Compared with four general clustering frameworks: DBSCAN, ST-DBSCAN, OPTICS and ST-OPTICS, we demonstrate that our method performs better than other clustering methods by evaluated on two internal cluster validity measures. Finally, we explore route choosing strategy according to the travel time of different trajectories and discover some abnormal trajectories.

CICTP