Authors: Miao Gao; Guoyou Shi
Addresses: Navigation College, Dalian Maritime University, Room 313-1, Nautical Building, Ling Hai 1# Road, High-tech Zone, Dalian 116026, Liaoning, China ' Navigation College, Dalian Maritime University, Room 313-2, Nautical Building, Ling Hai 1# Road, High-tech Zone, Dalian 116026, Liaoning, China
Abstract: The original automatic identification system (AIS) data are so large that they cannot be directly applied to learning and training, and the collision avoidance data must be filtered, identified, and extracted. AIS data from the Laotieshan channel in Dalian port, China, are used as raw data to identify successful cases of collision avoidance. Ship navigation statuses are screened according to AIS message codes. The improved density-based spatial clustering of applications with noise algorithm (DBSCAN) is used to cluster the four types of habitual routes of ship trajectory, with the rest of the data as candidate data for ship matching. Ship encounter situations are planned for 13 categories considering the ship light arc range and the requirements of the International Regulations for Preventing Collisions at Sea (COLREGs). The matched data utilise a sliding window algorithm for extracting ship navigation behaviour, which are then stored in the form of segmented ship trajectory unit sequences. This study suggests a new knowledge base of intelligent ship collision avoidance data, providing a novel method and theoretical guidance for future developments in ship collision avoidance methods.
Keywords: USV; ship collision avoidance; modelling; AIS; big data; extraction of behaviour feature; ship trajectory unit.
International Journal of Simulation and Process Modelling, 2020 Vol.15 No.1/2, pp.100 - 110
Received: 28 Aug 2018
Accepted: 31 Mar 2019
Published online: 22 Apr 2020 *