Maritime surveillance is a topic of significant strategic importance for both civil and security applications. Traditional ship detection methods, relying on hull identification in satellite imagery, are constrained by the limited resolution of satellite sensors. These limitations hinder accurate detection and prevent the extraction of key information for monitoring, such as ship speed and heading, which are essential for comprehensive maritime surveillance. The UEIKAP project, short for Unveil and Explore the In-depth Knowledge of Earth Observation Data for Maritime Applications, funded by the Italian Ministry of University and Research under the Research Projects of National Interest (PRIN), ending in early 2026, provides a solution to these limitations by shifting the detection focus to ship wakes. The developed framework is based on a multi-modal AI approach combining several sources of input data such as multi-spectral and radar satellite imagery, respectively from Sentinel-2 and Sentinel-1 platforms, local meteo-marine measurements, obtained from the ERA5 and CMEMS datasets, and Automatic Identification System (AIS) data for validation. Moreover, a method for generating synthetic wake imagery was developed to augment the datasets used to train the classification and detection models. Regarding the models, a framework based on the cascade application of several Convolutional Neural Networks (CNN) is used to first classify sea clutter and then to detect the wakes. The appearance of the wakes and their relative geometry with respect to the hull is used to estimate the speed, size, and heading of the detected vessels. Validation campaigns have been conducted in the Venice Lagoon (July 2024), and are ongoing in the Gulfs of Naples and Genoa (scheduled from May to October 2025). The results of the framework and of the validation campaigns are presented in this manuscript, detailing the methods applied for the development of project UEIKAP and during the test campaigns.
Ueikap Project: Results, Main Outcomes and Future Advances
Vernengo G.;Villa D.;Petacco N.;
2025-01-01
Abstract
Maritime surveillance is a topic of significant strategic importance for both civil and security applications. Traditional ship detection methods, relying on hull identification in satellite imagery, are constrained by the limited resolution of satellite sensors. These limitations hinder accurate detection and prevent the extraction of key information for monitoring, such as ship speed and heading, which are essential for comprehensive maritime surveillance. The UEIKAP project, short for Unveil and Explore the In-depth Knowledge of Earth Observation Data for Maritime Applications, funded by the Italian Ministry of University and Research under the Research Projects of National Interest (PRIN), ending in early 2026, provides a solution to these limitations by shifting the detection focus to ship wakes. The developed framework is based on a multi-modal AI approach combining several sources of input data such as multi-spectral and radar satellite imagery, respectively from Sentinel-2 and Sentinel-1 platforms, local meteo-marine measurements, obtained from the ERA5 and CMEMS datasets, and Automatic Identification System (AIS) data for validation. Moreover, a method for generating synthetic wake imagery was developed to augment the datasets used to train the classification and detection models. Regarding the models, a framework based on the cascade application of several Convolutional Neural Networks (CNN) is used to first classify sea clutter and then to detect the wakes. The appearance of the wakes and their relative geometry with respect to the hull is used to estimate the speed, size, and heading of the detected vessels. Validation campaigns have been conducted in the Venice Lagoon (July 2024), and are ongoing in the Gulfs of Naples and Genoa (scheduled from May to October 2025). The results of the framework and of the validation campaigns are presented in this manuscript, detailing the methods applied for the development of project UEIKAP and during the test campaigns.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
