Maritime traffic monitoring has long been a core topic in remote sensing, with major implications for security, environmental protection, and the monitoring of illegal activities. Current Earth Observation (EO) missions, while supplying valuable data for ship detection, are limited by their specific temporal and spatial coverage constraints. Activities such as illegal fishing, smuggling, human trafficking, and the deliberate sinking of unregistered vessels underscore the need for tools capable of continuous, large-scale surveillance of critical areas like the Mediterranean basin. The SPECTRE project (ShiP dEteCTion based on aRtificial intElligence, agreement no. 2023-4-E.0), funded by the Italian Space Agency (ASI) under the RESEARCH DAY ASI 2020 program, addresses this limitation by developing a ship detection framework that integrates multi-mission multi-frequency satellite data; this approach allows for higher revisit frequency and enable vessel tracking over broad maritime regions. A key element of the project is the creation and characterization of the first COSMO-SkyMed, first generation (CSK) and second generation (CSG), ship detection dataset. Images with multiple polarizations and resolutions are cropped and annotated via automated workflows, with vessel positions aligned to interpolated Automatic Identification System (AIS) tracks. The resulting dataset provides standardized image crops, polygonal annotations, and rich metadata suitable for training state-of-the-art detection networks. The performance of a YOLOv11 on the COSMO-SkyMed dataset is reported in the manuscript.
A COSMO-SkyMed Dataset For Maritime Monitoring / Galario, A., Daniela Graziano, M., Mazzeo, A., Carmen Cristofano, A., Renga, A., Ravellino, F., Pisacane, V., Aurigemma, R., Raimondo Focone, M., Lo Russo, A., Volpe, A., Tapete, D., Virelli, M.. - (2026). (EUSAR 2026 Baden-Baden, Germany ) [10.30420/456729140].
A COSMO-SkyMed Dataset For Maritime Monitoring
Andrea Galario;Maria Virelli
2026
Abstract
Maritime traffic monitoring has long been a core topic in remote sensing, with major implications for security, environmental protection, and the monitoring of illegal activities. Current Earth Observation (EO) missions, while supplying valuable data for ship detection, are limited by their specific temporal and spatial coverage constraints. Activities such as illegal fishing, smuggling, human trafficking, and the deliberate sinking of unregistered vessels underscore the need for tools capable of continuous, large-scale surveillance of critical areas like the Mediterranean basin. The SPECTRE project (ShiP dEteCTion based on aRtificial intElligence, agreement no. 2023-4-E.0), funded by the Italian Space Agency (ASI) under the RESEARCH DAY ASI 2020 program, addresses this limitation by developing a ship detection framework that integrates multi-mission multi-frequency satellite data; this approach allows for higher revisit frequency and enable vessel tracking over broad maritime regions. A key element of the project is the creation and characterization of the first COSMO-SkyMed, first generation (CSK) and second generation (CSG), ship detection dataset. Images with multiple polarizations and resolutions are cropped and annotated via automated workflows, with vessel positions aligned to interpolated Automatic Identification System (AIS) tracks. The resulting dataset provides standardized image crops, polygonal annotations, and rich metadata suitable for training state-of-the-art detection networks. The performance of a YOLOv11 on the COSMO-SkyMed dataset is reported in the manuscript.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


