摘要
由于在传统的CPU平台上进行计算耗时量大,一方面由于SAR回波数据量大,另一方面成像算法复杂。而处理核心众多则是GPU一大优势,适合独立并行结构算法的加速。文中借助GPU强大的浮点运算和高度的并行处理能力,将SAR成像中ECSA算法在GPU上进行了验证,并得出较好的效果,综合(计算时间+IO时间)加速比有了一定提高。在高分辨率星载SAR成像领域中,CUDA-GPU的运用将是未来发展方向,文中给出了利用GPU处理星载SAR数据可行的初步结论,为进一步优化程序奠定了基础。
Computation on traditional CPU platforms consumes large amount of time owing to large SAR echo data and imaging algorithm complexity. GPU has the great advantage of many processing cores suitable for the acceleration of independent parallel structure algorithms. With the powerful floating point and parallel processing capabilities of GPU, the ECSA SAR imaging algorithm is verified on the GPU with good results, and the comprehensive （ computation time ＋ IO time） speedup has been raised. Since CUDA-GPU utilization will be the future of high reso-lution spaceborne SAR imaging, the paper gives a preliminary indication of the feasibility of GPU to process spaceborne SAR data, which lays the foundation for further procedure optimization.
出处
《电子科技》
2013年第10期29-32,35,共5页
Electronic Science and Technology
