基于GPU的毫米波雷达近场阵列成像技术研究  

Research on near field array imaging in the millimeter band based on GPU

作　　者：杨啸宇 高敬坤 邓彬[1] 王宏强[1] 秦玉亮[1] Yang Xiaoyu;Gao Jingkun;Deng Bin;Wang Hongqiang;Qin Yuliang(Department of Intelligent Perceptual,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学智能感知系

出　　处：《电子测量技术》2018年第11期15-19,共5页Electronic Measurement Technology

摘　　要：距离徙动算法（RMA）可以对整个区域基于散射点模型而不加其余近似条件完成无几何形变的完全聚焦,在原理上是毫米波近场均匀阵列成像体制的最优成像算法。RMA在雷达近场阵列成像中应用广泛,但是计算量大,运用图形处理器（GPU）并行程序设计可提高计算速率。基于GPU在雷达近场阵列成像中用Python编译器Numba可加快RMA。NVIDIA GTX 850M和Intel（R）Core（TM）i5-4200H CPU上的实测数据表明,与传统RMA比,所提方法可以提高计算速度2.89倍。因此运用GPU来提高毫米波雷达近场阵列RMA成像速率是可行的。RMA can achieve the full focus without geometric distortion based on the scatter point model and without other approximate conditions in the whole region.In principle,it is the best imaging algorithm of millimeter wave near field uniform array imaging system.Range migration algorithm（RMA）is widely used in radar near field array imaging,but with large amount of computation. The Graphic Processing Unit（abbreviation GPU）parallel programming can improve the computation speed.Based on GPU,the python compiler Numba can speed up the imaging of the RMA algorithm in the radar near field array.The measured data on NVIDIA GTX 850 Mand Intel（R）Core（TM）i5-4200 H CPU shows that this method can increase the computing speed by 2.89 times compared with the traditional RMA.Therefore,it is feasible to use GPU to improve the near-field array RMA imaging rate of millimeter wave radar.

关 键 词：RMA Numba 近场阵列成像 GPU 

分 类 号：TN95[电子电信—信号与信息处理]