A Deep Learning-based Multi-Modal Approach for Wake Detection

The detection of 'dark' ships, vessels operating without Automatic Identification System (AIS) transponders, is a crucial challenge for maritime security and international law enforcement. Traditional methods based on optical and Synthetic Aperture Radar (SAR) imagery face limitations, particularly in detecting small, fast-moving vessels under complex environmental conditions. The UEIKAP project, funded by the Italian Ministry of University and Research under the Research Projects of National Interest (PRIN), explores a novel approach to ship detection by analyzing their wakes, which provide key information about vessel activity.This paper presents a multi-modal Deep Learning (DL)-based wake detection framework that integrates optical and SAR satellite imagery, AIS data, meteo-marine measurements, and synthetic wake models. The system integrates Convolutional Neural Networks (CNNs) and contextual environmental data to enhance classification accuracy. The first experimental campaign, conducted in the Venice lagoon in July 2024, provided initial validation of the method, confirming the dependency of wake visibility on local meteo-marine conditions. Future campaigns in the Gulfs of Genoa and Naples will further refine the methodology.By combining advanced DL techniques with multidisciplinary data fusion, the UEIKAP project aims to improve maritime domain awareness, offering a robust tool for vessel monitoring and regulatory enforcement. This paper describes the operational planning of the validation campaigns set to take place during Spring-Summer 2025. It also provides a general overview of the project's framework, giving details regarding the processing pipeline, the involved algorithms, the training setup, the data, and some preliminary results.