Content and Channel Aware Adaptive Semantic Communications for Satellite-Integrated Internet

Efficient onboard compression and robust transmission of massive remote sensing images (RSI) are critical in the upcoming satellite-integrated Internet, which serve as the key enabler for the evolution of integrated remote sensing and communications. To address the challenges posed by the constrained onboard resources and time-varying satellite-to-ground channels, we propose a content and channel aware adaptive semantic communication (CASC) framework. Our CASC features a semantic encoder equipped with a full-dimension perception module (FPM) for extracting and preserving critical features from RSI. Concurrently, a shadowed-Rician channel module (SRCM) is designed to model and map the multi-level fading characteristics. These components are integrated by a weight-divided adaptive policy network (WAPN), which jointly considers the encoded RSI features from the FPM, and the estimated fading level from the SRCM. The WAPN dynamically optimizes the compression rate, enabling flexible and resilient RSI transmission and reconstruction. Simulation results demonstrate that our CASC framework only has 40% parameters compared to the benchmark schemes, and achieves robust transmission performance over multi-level SR channel fading, in terms of peak signal-to-noise ratio, structural similarity index measure and learned perceptual image patch similarity.