Probability-Based Ship Encounter Classification Using AIS Data

With the rapid development of the maritime industry, the traffic density of waterways has expanded significantly. Rising concern about safety becomes an important driving factor for developing an intelligent decision support system. In order to enhance situation awareness, the classification of ship encounters is needed. Considering the importance of encounter classification to maritime safety, this paper proposes a novel framework for the real-time classification of encounter types based on Automatic Identification System (AIS) data. First, encounter types are categorized based on expertise; then, the important features are used to construct a probability-based model. After that, sensitivity analysis (SA) is utilized to optimize the probability-based model by means of optimal feature selections. Experiments are conducted using real AIS data collected at the ferry crossing in Oslofjord. Furthermore, through case studies of the ferry and crossing traffic, the method is proven to be promising in effectiveness.