QUASI-TRACKING OF DARK SHIPS WITH AN AI-BASED MULTI-MISSION MULTI-FREQUENCY APPROACH

Maritime security faces persistent threats from illicit activities including poaching, smuggling, and human trafficking.”Dark” ships, vessels that do not transmit a valid Automatic Identification System (AIS) signal, exacerbate these challenges by evading detection. The SPECTRE project, funded by the Italian Space Agency (ASI) (with agreement n. 2023-4-E.0), addresses this issue through a multi-mission multi-frequency approach integrating data from Sentinel-1, Sentinel-2, COSMO-SkyMed, and the upcoming IRIDE mission to enhance ship detection and quasi-tracking capabilities. Using state-of-the-art models based on YOLOv11 for Synthetic Aperture Radar (SAR) and on Mask R-CNN for optical data, SPECTRE combines deep learning (DL) techniques with super-resolution methods to improve detection accuracy, particularly for small vessels. A novel spatial-temporal matching module links multi-mission observations to reconstruct ship trajectories, enabling extended monitoring beyond single-satellite limitations. Preliminary results show robust performance in ship detection and trajectory reconstruction on specifically annotated data, highlighting the feasibility of the approach. While challenges such as sensor heterogeneity and environmental variability remain, SPECTRE’s iterative development process ensures adaptability. By late 2025, the project aims to deliver a comprehensive platform for real-time maritime surveillance, enhancing security and compliance in critical regions throughout the Mediterranean Sea.