基于GPU的相位干涉仪FX鉴相算法

针对相位干涉仪测向系统对于大量高速实时信号的处理需求, 设计了基于图形处理单元(graphic processing unit, GPU)的频域互相关(简称为FX)鉴相算法, 完成了相应的并行程序设计, 进行了实时数据的测试验证。为充分发挥GPU强大的浮点运算能力和并行数据处理能力, 将涉及大量并行高速数据计算的核心鉴相算法加载在GPU中, 实现了高速并行数据的相关处理和相位提取; 利用中央处理器(central processing unit, CPU)完成了数据调度、分发和简单的数据处理功能。实验测试结果表明, 在较好地保证鉴相精度的条件下, 本文设计的基于GPU的鉴相算法, 其数据处理速度是基于CPU平台的140倍左右, 鉴相速度明显提升, 较为圆满地实现了实时性、可靠性和准确性的设计初衷。

FX phase identification algorithm for GPU-based phase interferometer

Aiming at the processing requirements of a large number of high-speed and real-time signals in the phase interferometer direction-finding system, a FX phase identification algorithm based on graphics processing unit (GPU) is designed. The corresponding parallel programming is completed, and the real-time data are tested and verified. In order to give full play to the powerful floating-point computation and parallel data processing capability of the GPU, the core phase detection algorithm which involves a large number of parallel high-speed data calculation is loaded into GPU, the correlation processing and phase extraction of high-speed parallel data are realized. At the same time, central processing unit (CPU) is used to complete functions of data scheduling, distribution, and simple data processing. The experimental results show that under the condition of guaranteeing the phase-detection accuracy, the data processing speed of the GPU-based phase detection algorithm designed is about 140 times that based on CPU platform. The phase-detection speed is obviously improved, and the design intentions of real-time, reliability, and accuracy are realized satisfactorily.