The distributed imaging processing method of space-borne SAR based on embedded GPU

With the growth of space-borne SAR echo data, the pressure on satellite to ground data transmission is increasing. Therefore, real-time imaging processing of space-borne SAR on-orbit becomes an urgent need. A data processing method of distributed embedded GPU based on Chirp Scaling (CS) imaging algorithm is proposed in this paper. We use multiple embedded GPUs to process large-scale space-borne SAR data jointly. Unified Memory is used to allocate memory to reduce data transmission between the CPU and GPU, and the allocated memory is reused to solve the limited computing resources of a single embedded GPU. In addition, CUDA streams are used to overlap data transmission and data processing time to reduce communication delays between embedded GPUs. The distributed processing method proposed in this paper can effectively reduce a single satellite's computing and storage resources and improve the full-orbit utilization of a single satellite. Multiple NVIDIA Jetson Nanos are used to build the simulation system, and the system's feasibility was verified.