Mining maritime traffic conflict trajectories from a massive AIS data

The growing volume of maritime traffic is proving a hindrance to navigational safety. Researchers have sought to improve the safety of maritime transportation by conducting statistical analysis on historical collision data in order to identify the causes of maritime collisions. However, this approach is hindered by the limited number of incidents that can be collected in a given area over a given period of time. Automatic Identification System (AIS) has made available enormous quantities of maritime traffic data. Trajectory data are collected through the electronic exchange of navigational data among ships and terrestrial and satellite base stations. Due to a massive AIS data of recording ship movement, such data provide great opportunity to discover maritime traffic knowledge of movement behavior analysis, route estimation, and the detection of anomalous behaviors. Our objective in this paper was to identify potential between-ship traffic conflicts through the discovery of AIS data. Traffic conflict refers to trajectories that could lead to a collision if the ships do not take any evasive action. In other words, conflicting trajectories can be treated as a near-collision cases for analysis. The prevention of collisions requires an efficient method by which to extract conflicting trajectories from a massive collection of AIS data. To this end, we developed a framework CCT Discovery that allows the automated identification of clusters of conflicting trajectories (CCTs) from AIS data without supervision. Experiments based on real-world data demonstrate the efficacy of the proposed framework in terms of accuracy and efficiency. For improvement in the navigational traffic safety, the discovered data of conflict trajectory can contribute to numerous applications, such as collision situation awareness and prediction, anti-collision behaviors modeling and recommendation, and conflict area analysis for maritime traffic flow management.