Detection of Abnormal Vessel Behaviours Based on AIS Data Features Using HDBSCAN+

Abstract

Achieving maritime security is challenging due to the vastness and complexity of the domain. Monitoring all Achieving maritime security is challenging due to the vastness and complexity of the domain. Monitoringall vessels that use this medium is humanly impossible but is needed for law enforcement. This paper proposes amachine learning solution based on HDBSCAN+ to classify the movements of vessels into ‘normal’ or ‘abnormal’.This classification reduces the number of vessels that have to be monitored by law enforcement agencies to amanageable size. To date, AIS is the primary source of information that can represent vessel movements andenable the detection of maritime anomalies. The proposed model uses latitude, longitude, type of vessel, courseand speed as features of the AIS data for analysis. The performance of the proposed model is validated against the marine incidents reported by Information Fusion Centre-Indian Ocean Region (IFC-IOR). The proposed model has successfully detected the incidents reported by IFC-IOR.


Visualisation of AIS data points (latitude & longitude) from 05 August to 09 August 2021.
… 


Visualisation of the HDBSCAN+ Clusters, based on AIS Position of Vessels (After grouping the data based on type of vessel).
… 


Visualisation of the 'Normal' and 'Abnormal' Points after Classifying them with HDBSCAN + based on the 'type of vessel', AIS position, Cog and Sog (the points shown in black are noise or abnormal, and the points shown in other colours are normal points).
… 


A summary of the results obtained from the proposed model
… 

 

Description