A Novel Method for Vessel Trajectory Clustering Based on Three-Dimensional Triangulation Division Algorithm

Automatic identification system (AIS) performs vital functions in vessel collision avoidance and maritime security. The large-scale vessel trajectory data recorded by AIS has been facilitating vessel trajectory pattern mining, of which vessel trajectory clustering is a fundamental and indispensable approach. The precise and efficient measurement of trajectory similarity serves as the cornerstone of large-scale trajectory clustering. Most ship trajectory similarity measurement algorithms have issues of high time complexity, poor robustness and inability to distinguish the trajectories in opposite directions. This paper presents an improved measurement named three-dimensional Triangulation Division (3TD), which is based on area division in three-dimensional space with time axis incorporated, to ameliorate those deficiencies. In addition, a novel vessel trajectory clustering method is proposed by combining the 3TD algorithm with DBSCAN algorithm. AIS trajectory data in Yangshan Port area has been used to make experimental dataset. Experimental results indicate that the improved method in this paper has superior performance compared with some other existing methods in terms of clustering accuracy, robustness and computational efficiencies.