MARITIME TRAFFIC ANOMALY DETECTION FROM AIS SATELLITE DATA IN NEAR PORT REGIONS
Maritime traffic monitoring is an important aspect of safety and security in close-to-port operations. While there is a large amount of data with variable quality, decision makers need reliable information about possible situations or threats. In this thesis, we propose a two-component maritime traffic anomaly detection model. First, it extracts normal ship trajectory patterns using, besides ship tracing data, the publicly available IMO Rules. The main result of clustering is a set of generated lanes that can be mapped to those de ned in the IMO directives. Then, we show how the second anomaly detection component detects anomalous navigational behaviors based on three specialized division distances with the clusters. It decides for each point if the vessel is anomalous, considering longitude, latitude, direction and speed. This point-based approach is applicable for real-time AIS (Automatic Identification System) surveillance; it is also feasible for analysts to set their own threshold for labeling whole trajectories.