The UEIKAP project (Unveil and Explore the Indepth Knowledge of Earth Observation Data for Maritime Applications) introduces a novel framework for detecting noncooperative vessels through the analysis of ship wakes in satellite imagery. Leveraging a multi-modal approach, the system integrates Sentinel-1 synthetic aperture radar (SAR) and Sentinel2 optical data, contextual meteo-oceanographic variables, AIS signals, and synthetic wake simulations. A deep learning pipeline, comprising a YOLOv11x-based segmentation model, a sea-state classification module, and a post-processing block for motion estimation, has been developed and validated across a series of experimental campaigns. These include controlled trials in the Venice Lagoon (July 2024) and real-world deployments in the Gulfs of Naples and Salerno (May-June 2025). Results show consistent wake detection across optical modalities, with strong alignment between model outputs and AIS records. A final campaign scheduled for September 2025 will further assess performance under diverse sea states and vessel classes. UEIKAP aims to provide a robust, extensible tool for maritime surveillance and environmental monitoring, with open-access datasets and tools scheduled for release by the end of 2025.
Evaluation of a Multi-Modal Wake Detection Framework Using Satellite Data / Graziano, M.D., Vernengo, G., Bonaldo, D., Cristofano, A.C., Mazzeo, A., Renga, A., Villa, D., Petacco, N., Scarpa, G.M., Braga, F., Vavasori, P., Menegon, S., Fadini, A.. - (2025), pp. 527-531. (IEEE International Workshop on Metrology for the Sea; Learning to Measure Sea Health Parameters, MetroSea 2025 ita 2025) [10.1109/MetroSea66681.2025.11245743].
Evaluation of a Multi-Modal Wake Detection Framework Using Satellite Data
Graziano M. D.;Cristofano A. C.;Mazzeo A.;Renga A.;
2025
Abstract
The UEIKAP project (Unveil and Explore the Indepth Knowledge of Earth Observation Data for Maritime Applications) introduces a novel framework for detecting noncooperative vessels through the analysis of ship wakes in satellite imagery. Leveraging a multi-modal approach, the system integrates Sentinel-1 synthetic aperture radar (SAR) and Sentinel2 optical data, contextual meteo-oceanographic variables, AIS signals, and synthetic wake simulations. A deep learning pipeline, comprising a YOLOv11x-based segmentation model, a sea-state classification module, and a post-processing block for motion estimation, has been developed and validated across a series of experimental campaigns. These include controlled trials in the Venice Lagoon (July 2024) and real-world deployments in the Gulfs of Naples and Salerno (May-June 2025). Results show consistent wake detection across optical modalities, with strong alignment between model outputs and AIS records. A final campaign scheduled for September 2025 will further assess performance under diverse sea states and vessel classes. UEIKAP aims to provide a robust, extensible tool for maritime surveillance and environmental monitoring, with open-access datasets and tools scheduled for release by the end of 2025.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
