Leveraging SAR Satellites and AI for Real-Time Dark Vessel Detection and Maritime Surveillance

Maritime surveillance plays a pivotal role in safeguarding coastal security, economic stability, and environmental resources. “Dark vessels,” which intentionally switch off Automatic Identification System (AIS) transponders, pose significant threats through illicit fishing, smuggling, and other unlawful activities. To address these concerns, we propose an integrated framework that fuses Synthetic Aperture Radar (SAR) imagery with AIS data for near real-time vessel detection and velocity estimation. Our approach comprises a deep-learning detector (YOLOv4), a Doppler-based phase-refocusing module that corrects azimuth defocusing for moving targets, and a system for flagging dark vessels lacking AIS signals. Evaluations on multi-sensor SAR datasets (e.g., Sentinel-1, ICEYE) show an F1-score above 85%, along with reduced geolocation errors thanks to our motion compensation. This paper discusses the end-to-end pipeline-from data acquisition and preprocessing to vessel detection, classification, dark-vessel identification, and velocity-based refocusing.