基于CPU-GPU异质架构的SAR实时成像加速

对于大规模SAR回波数据，单纯依靠CPU实现实时成像处理存在较大挑战。近年来出现的GPU图像处理单元为SAR成像处理提供了一个理想的处理平台。本文结合GPU和CPU的特点，利用基于CPU-GPU的异构协同处理平台来加速SAR成像算法。同时，为了充分发挥该异质架构的性能，文中提出了一种改进的基于子孔径架构的成像算法，并在此架构上对超大规模SAR回波数据进行了成像处理，取得良好的成像效果和加速比。

Acceleration of SAR Real-time Imaging based on the Heterogeneous CPU-GPU Architecture

Traditionally, SAR Imaging is achieved by CPU alone, which is too time-consuming to ensure real-time imaging. Recently, the rapid increase in the performance of GPU has made it into a compelling platform for SAR Imaging. In this paper, the hybrid CPU/GPU parallel system is presented to accelerate SAR Imaging Algorithm. Meanwhile, in order to fully exploit the potential capacity of the heterogeneous computing platform, the paper proposes a novel Sub-aperture Imaging Algorithm to achieve real-time imaging.