Research on real-time imaging method of spaceborne SAR based on embedded GPU

Synthetic aperture radar (SAR) as an all-time, all-weather high-resolution imaging radar, can quickly and conveniently acquire the spatial distribution characteristics of a wide range of observation areas by carrying a satellite-borne platform, and it is widely used in military and civilian fields. The data processing mode of a traditional spaceborne SAR system is that the satellite obtains the echo data and transmits it to the ground handling equipment for processing. This processing mode has obvious disadvantages, which greatly reduces the timeliness of acquiring ground information. In this paper, an on-board SAR real-time imaging method based on the embedded GPU is proposed, which can be called improved GPU-CS algorithm. This method combines GPU parallel processing with CS algorithm, adopts zero-copy technology, optimized phase data construction method, GPU-shared memory transpose, and other improved methods to realize fast imaging. Imaging validation was performed on the Jetson AGX Xavier. The improved GPU-CS algorithm is 2~3 times faster than the traditional GPU-CS algorithm and provides method support for on-board SAR real-time imaging.