DenseRefrception: A Densely Connected Perception Framework for Passive Retrieval of Ocean Refractive Environments via GPS Scattered Signals

Accurate ocean ducting environments sensing is crucial for radio communication. Conventionally, ducting profile inversion relied heavily on radar sea returns. Though feasible, this active approach suffers from operational drawback, which severely impedes practical deployment. Global positioning system (GPS) has revolutionized ocean monitoring through its inherent global coverage, all-weather capability, and high precision. GPS signals scattered from the sea surface provide unique yet underutilized information on atmospheric environments. Although duct-induced modulations encode key ducting parameters, direct extraction from these signals poses intractable nonlinear inverse problems due to complex signal-refraction interactions. Leveraging the unparalleled capability of deep learning to approximate highly complex, nonlinear mappings, this study presents aDensely Connected Perception Framework for Atmospheric Ducting Retrieval via GPS Scattered Signals(DenseRefrception). The core breakthrough resides in constructing a sophisticated architecture. This framework innovatively integratesDense Connectivityto extract discriminative features from scattered GPS signal power, while incorporating multi-head attention and fully-connected networks to establish high-dimensional nonlinear mappings. Through systematic training of a unified multi-dimensional duct parameter learning mechanism, it precisely constructs highly nonlinear functional relationships. By transforming passively collected GPS signals into actionable refractivity intelligence, the DenseRefrception represents a shift towards persistent and clandestine monitoring of the oceanic environment, significantly enhancing situational awareness for maritime platforms.