Ship encounter azimuth map division based on automatic identification system data and support vector classification

Abstract Currently, the division of encounter situations and collision avoidance decisions both depend on the individual subjective judgment of officers under conditions of extraordinary complexity and randomness. Ambiguities and contradictions are present among the existing quantifications of azimuth division from the International Regulations for Preventing Collisions at Sea (COLREGS), radar collision avoidance diagrams, and expert questionnaire results. At present, there is no unified and practical division model for the variety of azimuth divisions encountered by ships. With the development of intelligent ship technology, the realization of maritime autonomous surface ships is possible. However, more obscure problems must be accurately defined. Moreover, the requirements for an accurate division of the ship encounter situation in maritime accident analysis are becoming more intense. Additional requirements have been imposed on the division of azimuth, and ship encounters have been quantified into multiple features for machine learning. In this study, automatic identification system data near Zhoushan Port were used to reproduce the relative motion process of ships, and extract the meeting position of the ship and the corresponding actual avoidance behavior. By combining the requirements for the light range in COLREGS and support vector classification to supervise and learn the actual meeting data, a map of the ship encounter azimuth division was constructed. The map can serve as an accurate numerical basis for the division of marine encounter situations, maritime accident responsibility division, and intelligent ship collision avoidance decisions.