Ship abnormal behaviour detection based on AIS data at the approach to ports

This study proposes a novel framework for abnormal ship behaviour detection, extracting traffic characteristics in port approach waters from collected Automatic Identification System (AIS) data and establishing corresponding anomaly indicators to measure the degree of abnormal vessel behaviour. The method involves shipping route modelling, which consists of route centreline definition and boundaries quantification. Then, the constructed model is applied to the extraction of traffic position characteristics and the generation of GraphSAGE-based dynamic transportation patterns. Subsequently, the optimal distribution-based abnormal indicators of position, Speed Over Ground, and Course Over Ground are developed with the extracted position characteristics and generated dynamic patterns. The effectiveness of the proposed study is validated with the outbound behaviour abnormal indicators of Leixões port in Portugal. The results show objectively, effectively, and interpretably that the method quantifies vessel behaviour anomalies based on local traffic characteristics.