Multi-Task Driven Semantic Communication for Satellite Imagery

In the space-air-ground integrated networks of the sixth generation (6 G) systems, many satellite imagery need to be transmitted from the satellite to the ground with high resolution for further processing. However, it faces the challenges of limited available bandwidth and the poor channel conditions of satellite-to-ground links. In this paper, leveraging the benefits of semantic communication for efficient transmission under limited bandwidth and low signal-to-noise ratio (SNR) conditions, a joint Preprocessing and Multi-task driven Semantic Communication (PMSC) system for satellite imagery transmission is proposed. To efficiently use the limited bandwidth, we propose a region of interest (ROI) based preprocessing method, which focuses on only relevant regions that will be encoded into semantic information, and processes the ROIs that are pivotal for the tasks. Moreover, we formulate a multi-task driven semantic communication system with a universal joint semantic-channel encoder and distinct decoding processes, making the received semantic features of satellite imagery can be accurately and effectively utilized for different applications. The simulation results demonstrate that the proposed PMSC system has better performance in enhancing reconstruction and classification in the target regions of interest at the same compression ratio, especially under low SNR conditions.