Automated wake detection from marine satellite imagery: lessons learned, methods, and perspectives from the UEIKAP project

Monitoring non-cooperative or poorly surveyed maritime traffic remains challenging when Earth Observation (EO) systems rely mainly on direct hull detection, especially for small vessels and cluttered marine scenes. This paper presents the main lessons learned from the Unveil and Explore the In-depth Knowledge of Earth Observation Data for Maritime Applications (UEIKAP) project, which investigated ship wakes as alternative, information-rich observables for vessel detection and characterization in satellite imagery. The proposed framework combines Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 multispectral imagery, Automatic Identification System (AIS) data for supervision and validation, meteo-marine contextual information, and synthetic wake simulations generated through numerical modeling. Deep learning (DL) segmentation models are used to detect wakes and, in SAR imagery, ship hulls. A dedicated post-processing chain then extracts wake geometry and derives pseudo-AIS variables, including course-related quantities and speed estimates from azimuth-shift effects. The results show that wake-based monitoring is technically feasible and can extend satellite-based maritime awareness beyond target localization alone. A central lesson of the project is that optical imagery currently provides the most reliable wake delineation under favorable visibility conditions, whereas SAR offers unique potential for kinematic inference but is much more sensitive to environmental forcing and acquisition geometry. The validation campaigns further show that wake detectability is strongly conditioned by meteo-marine context and vessel characteristics, making environmental awareness a key requirement for operational use. Overall, this paper presents the UEIKAP multimodal framework together with open-source Sentinel-1 and Sentinel-2 wake datasets, and defines the main methodological steps required to improve the robustness and operational applicability of wake-based monitoring for maritime surveillance and sea-use monitoring.