Integrated satellite–AIS–UAV framework for suspected non-cooperative vessel screening in complex estuarine waters

Maritime surveillance in complex estuarine waters is constrained by incomplete vessel self-reporting, uncertain satellite-AIS association, and difficulties in converting remote-sensing observations into executable verification tasks. Existing satellite-AIS monitoring improves coverage but often lacks integrated treatment of association uncertainty and verification feasibility. This study addresses this operational gap through an integrated satellite–AIS–UAV framework for suspected non-cooperative vessel screening in the Pearl River Estuary using GF-6 PMS imagery, AIS trajectory data, oriented vessel detection, and low-altitude UAV-response analysis. The proposed FE-P2E detector improved vessel localization, increasing Precision, F1-score, and mAP50-95 by 8.0%, 7.1%, and 16.0%, respectively, relative to YOLO11n-OBB. FE-P2E remained competitive on HRSC2016, supporting its adaptability beyond the GF-6 scene. Association analysis identified 93 of 334 deduplicated detections (27.84%) as suspected candidates without reliable AIS associations. At a 300 m verification radius, the static current strategy achieved 98.21% success for stationary-like targets, whereas the heading local context strategy increased success for underway targets from 1.65% to 33.88% and reduced median handoff error by 62.8%. Perturbation analysis showed that positional errors affected AIS assignment and UAV handoff more strongly than temporal offsets. These results demonstrate that maritime verification requires coordinated uncertainty control across detection, AIS association, and motion-aware response.