Autonomous Planning for On-Board Adaptive Remote Sensing Transmission

The enormous increase of in-orbit satellites, exemplified by mega constellations like Starlink, highlights the scarcity of on-board resources in Earth-imaging satellites, which struggle to meet the high demand and transmit large volume of remote sensing images generated due to limited computing hardware, battery capacity, and available time windows. In addition, the limited downlink bandwidth also makes the real-time transmission of high-resolution remote sensing images to the ground challenging. In this paper, an adaptive planning method is proposed for resolution allocation of image transmission, which considers both image content and dynamic downlink capacity. Aiming at onboard autonomy, regions of interest (ROIs) are first extracted accurately in a high speed by utilizing multi-scale hybrid saliency mapping in combination with semantic refinement. In order to adapt to the dynamics of downlink bandwidth, the nonRoIs are proactively downsampled by optimizing the trade-off between data volume and the real-time prediction of transmission speed. The experimental results demonstrate that the proposed method can provide efficient remote sensing image transmission, significantly reducing the redundant data and alleviating the transmission pressure to ground stations within the constraints of limited on-board environments.