基于 GPU 的 SAR 回波仿真高效实现方法

摘要

Due to the low efficiency of synthetic aperture radar (SAR)raw data generation in distributed scene,a high-efficient SAR echo simulation method based on the modified concentric circles algorithm and the graphics processing unit (GPU)is proposed.Firstly,an improved concentric circles algorithm is proposed to a-void cases of low signal to noise ratio (SNR)when the traditional algorithm whose precision is low is used.Sec-ondly,the parallel processing of GPU is deeply optimized in order to lift the efficiency of the simulation and make full use of the parallel advantages in GPU kernel.Then,the proposed method adopts “thread extrapola-tion”to accumulate partial data and realize the whole data generation via “reduction adding”.Finally,simulation results validate the high precision and eficiency of the proposed approach.%针对合成孔径雷达（synthetic aperture radar，SAR）分布式场景回波仿真计算量巨大的问题，提出了一种结合改进的同心圆算法与图形处理器（graphics processing unit，GPU）技术的高效 SAR 回波仿真方法。首先，针对常规同心圆算法精度较低造成的图像信噪比低的问题，提出了一种改进的同心圆算法。其次，为了充分发挥 GPU 处理核之间的并行优势，对该算法的 GPU 并行处理进行了深度优化，进一步提升了仿真速度。具体方法是，根据并行度的高低设计核函数，确定了先采用“线程外推”实现部分目标回波的同心圆累加，再用“归约相加”实现所有目标回波的累加。最后，与常规 GPU 方法进行了实验对比，验证了所提方法的精确性和高效性。