PUBLICATION
A Reliable Remote Sensing-Based Framework for Vessel Detection
Type
Conference Paper
Year
2026
Authors
Abubaker Gaber
Stelios P. Neophytides
Dr.-Ing. Sebastian Heil
Nithinkumar Mirle
Michalis Mavrovouniotis
Georgios Kavallieratos
Georgios Spathoulas
Prof. Dr.-Ing. Martin Gaedke
Research Area
Event
Workshop on Machine Learning for Earth Observation
Published in
Machine Learning and Principles and Practice of Knowledge Discovery in Databases
ISBN/ISSN
978-3-032-19108-3
Download
Abstract
Reliability in maritime surveillance systems is vital to ensure con-sistent and effective monitoring of oceanic activities. As traditional land-based monitoring is limited in its applicability to vast and dynamic marine environments, synthetic aperture radar (SAR) satellites provide a dependable alternative, offering high-resolution, all-weather imaging capabilities. The growing threat of unregu-lated maritime behaviour, such as illegal fishing, unauthorised border crossings, and vessel concealment, has increased the demand for robust and automated detec-tion systems. Particularly, the identification of “dark vessels” that operate without automatic identification system (AIS) signals poses significant challenges. These vessels often engage in illicit activities while exploiting gaps in the current surveil-lance infrastructure. Therefore, reliable ship detection mechanisms that utilise only SAR imagery are crucial for enhancing situational awareness and maritime domain security. This study proposes a framework that enhances ship detection in SAR images by combining the outputs of multiple object detection algorithms to increase the reliability of the surveillance system. The framework operates not only on raw SAR images but also on their derived forms, such as filtered versions or colour-enhanced representations, to provide a more robust assessment of the presence of ships.
Reference
TBA
