Self-learning adaptive algorithm for maritime traffic abnormal movement detection based on virtual pheromone method

The paper deals with a newly designed self-learning adaptive algorithm enabling detecting any non-standard movement in marine traffic based on the bio-inspired virtual pheromone method. The algorithm detects non-standard vessel movements purely based on the common marine traffic patterns from the Automated Identification System. The proposed approach provides rapid self-learning and fast adaptation characteristics. We verified the algorithm´s accuracy in two modes, each of which incorporates different learning factors. The dataset for the verification of the proposed algorithm was provided by the marine traffic Automated Identification System of the Klaipeda seaport.