Vessel Tracking & Multi-Source Data Fusion

Detection tells you a ship is there. Fusion tells you who it is and, far more usefully, which ships are there without saying so. No single source manages that on its own, because every source has a blind spot the others cover. The work of this service is to bring all of them onto one timeline, one geometry and one identity, and then to report what is left over.

Four sensing layers go in. Self-reported positions come from AIS, terrestrial and satellite, and where the customer is an authority that holds them, from long-range identification and tracking reports and from fisheries vessel monitoring systems. Imaging detections come from the platform’s own radar and optical services, which between them cover night, weather and the small non-metallic hulls radar cannot see. Radio-frequency geolocation finds vessels by their emissions (navigation radar, VHF, satellite telephone) which works precisely when a transponder has been switched off. Night-time light detection finds light-luring fishing fleets that neither radar nor daylight optical reliably catch.

Two more layers turn positions into meaning. Registries resolve a detection into a named vessel: the international ship registry, national and EU fleet registers, the authorised-vessel and illegal-fishing lists maintained by regional fisheries bodies, sanctions designations, ownership and flag history, and port state control inspection records. Ancillary marine data supplies the context that makes a position interpretable, EMODnet for bathymetry, vessel density, wind farms, cables, pipelines and extraction sites; exclusive economic zones, territorial waters, protected areas, traffic separation schemes and port polygons for jurisdiction; and met-ocean fields from the Copernicus marine service for wind, waves, currents and sea ice. Two of these do real analytical work rather than decoration: bathymetry rules out positions a hull of a given draft physically cannot occupy, and sea state governs how much the radar could have missed, which is what turns a completeness statement from a disclaimer into a number.

The output is a traffic picture where every position carries its provenance, which source saw it, when, and with what confidence, and every detection is classified as matched, unmatched or ambiguous. The unmatched ones are the operational product.