Semantic behaviour modelling for threat detection in maritime security

Chen, L., Patino, L. ORCID: https://orcid.org/0000-0002-6716-0629, Boyle, J. ORCID: https://orcid.org/0000-0002-5785-8046, Markchom, T. ORCID: https://orcid.org/0000-0002-2685-0738 and Ferryman, J. (2026) Semantic behaviour modelling for threat detection in maritime security. Applied Soft Computing, 203 (Part B). 115968. ISSN 1872-9681 doi: 10.1016/j.asoc.2026.115968

Abstract/Summary

This work presents a semantic behaviour analysis framework designed for real-time maritime surveillance applications. The system takes input from object detection and tracking generated by heterogeneous sensors, such as ground-based visual and thermal cameras, high- and low-altitude remote sensors and Automatic Identification System (AIS). A zone-based soft computing approach is employed to detect behaviour events which automatically analyses calculated low-level trajectory data to identify individual and pairwise interactions between objects. These behaviour events are then translated into human-readable semantic descriptions using a domain-informed rule-based model. The proposed framework supports a variety of operational environments across land, sea or combined surveillance areas, with a particular focus on threat event detection near the coastline, including interactions between vessels, people, and vehicles. Extensive evaluation was performed using a selection of datasets including both collected real-world data and generated simulated data that mimic real-world maritime security scenarios near the coast. The results demonstrate the system’s ability to robustly and consistently detect semantically meaningful events across diverse real-time operational deployments and input data sources.

Item Type	Article
URI	https://centaur.reading.ac.uk/id/eprint/131166
Identification Number/DOI	10.1016/j.asoc.2026.115968
Refereed	No
Divisions	Science > School of Mathematical, Physical and Computational Sciences > Department of Computer Science
Publisher	Elsevier
Download/View statistics	View download statistics for this item