系统工程与电子技术

• 传感器与信号处理 • 上一篇    下一篇

基于GPU的SAR回波仿真高效实现方法

景国彬1,2, 张云骥1, 李震宇1,2, 孙光才1,2, 邢孟道1,2, 保铮1,2   

  1. 1. 西安电子科技大学雷达信号处理国家重点实验室, 陕西 西安 710071
    2. 西安电子科技大学信息感知技术协同创新中心, 陕西 西安 710071
  • 出版日期:2016-10-28 发布日期:2010-01-03

Efficient realiazation of SAR echo simulation based on GPU

JING Guo-bin1,2, ZHANG Yun-ji1, LI Zhen-yu1,2, SUN Guang-cai1,2,XING Meng-dao1,2, BAO Zheng1,2   

  1. 1. National Lab of Radar Signal Processing, Xidian University, Xi’an 710071, China; 2. Collaborative
    Innovation Center of Information Sensing and Understanding, Xidian University, Xi’an 710071, China
  • Online:2016-10-28 Published:2010-01-03

摘要:

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。

Abstract:

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 avoid cases of low signal to noise ratio (SNR) when the traditional algorithm whose precision is low is used. Secondly, 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 extrapolation” 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.