Design And Implementation Of Digital Twin System For On-orbit Processing On-brbit SAR

With the continuous development of satellite-based SAR technology,the amount of data obtained by SAR satellites is growing rapidly,which makes the satellite-based SAR system suffer from such problems as high pressure of satellite-ground data transmission and long system response time,and the traditional working mode of transmitting data to the ground for processing is difficult to cope with the current mission requirements.Therefore,the on-orbit processing of SAR data on board has received more and more attention.The development and testing costs of real on-board SAR in-orbit processing system are high,and in order to reduce the risk of system development,it is necessary to build a simulation system on the ground to verify the feasibility and effectiveness of the real system.At the same time,the on-orbit processing of satellite-based SAR requires high computational resources,and it is difficult to match the limited hardware resource environment on board,so it is necessary to optimize its processing steps and processing status with the help of numerical simulation methods to improve the operation efficiency.Therefore,the main work of this paper is to give a digital twin system for on-orbit processing of satellite-based SAR in a resource-constrained environment.The specific work is as follows:(1)A digital twin system for on-orbit processing of satellite-based SAR is constructed.The on-orbit processing system of satellite-based SAR is divided into three parts: data acquisition,data processing and information extraction.In this paper,the system is mapped into the virtual space according to the digital twin technology,and the data acquisition twin model,imaging processing twin model and image interpretation twin model are constructed.In terms of hardware platform,a low-power device is used to build the system runtime platform for deploying the model,which takes into account both low-power and low-latency characteristics and effectively simulates the hardware environment with limited computing resources on the star.Subsequent experiments show that the system can effectively realize the digital twin for the whole process of on-board SAR processing in orbit.(2)A joint computational storage optimization scheme for the digital twin system of on-board SAR in-orbit processing is designed.For the problems of limited onboard power,limited computing power and limited storage,the digital twin process of on-orbit processing is optimized in terms of storage and computation based on GPU parallel technology and deep model compression technology.The data acquisition process is optimized by using data dynamic chunking strategy,streaming technique,Fourier transform optimization,and data type conversion;the imaging processing process is optimized by using storage optimization,matrix transpose optimization,and phase factor multiplication optimization operation;the image decoding process is optimized by using a combination of channel pruning and knowledge distillation for detection model lightweighting.The experimental results show that after the storage optimization,the efficiency of data acquisition,imaging processing and image decoding is effectively improved,and the model storage volume is effectively compressed,which verifies the effectiveness of the optimization scheme.