A novel method for extracting near-misses from AIS data involving collision avoidance behaviour

Abstract

This paper presents a novel method for extracting near-misses involving collision avoidance behaviour based on automatic identification system data. Previous methods for near-misses extraction primarily focused on detecting collision risks from ship-to-ship encounter scenarios. The proposed method introduces avoidance behaviour as an additional factor for extracting near-misses. It further analyzes the subject of avoidance behaviour through the avoidance intention, aiming to extract near-misses where one ship's avoidance behaviour is prompted by another ship within the same ship-to-ship scenario. The method consists of three primary stages: (1) Detection of encounter scenarios with collision risk using the velocity obstacle method; (2) Identification of encounter scenarios with maneuvering behaviour based on course data; (3) Recognition of near-misses involving avoidance behaviour by estimating avoidance intention. Finally, the effectiveness and innovativeness of the proposed method are validated through a comparative experiment, and the influence of various parameters is thoroughly analyzed in an applied experiment.