相关干涉仪算法具有实现简单和精度高等优点,但将其运用于宽带测向时存在运算量较大等问题,文中针对该问题 提出一种基于GPU的宽带干涉仪测向算法实现,利用CUDA技术对传统相关算法进行改进,使之能够适应于GPU平台并 充分发挥图形处理器强大的浮点运算能力及其出色的并行执行性,进而使算法在执行速度上能获得极大提升。算法中采 用向量1范数并通过插值拟合得到来波方向的精确估计,仿真结果表明,算法在满足实时性要求的同时也能够保证较高 精度。相比于CPU 平台,算法的GPU 实现能够获得很高的时间加速比。

Correlative interferometer algorithm is known for its easy implementation and high accuracy, but its advantage is impaired due to large amount of calculation when it is applied to wideband situation. Concerning with this problem, a wideband correlative interferometer algorithm for direction finding based on GPU is described. CUDA is used to make adaptive improvement on the traditional correlative algorithm in the hope of exploiting GPU's tremendous computational horsepower as well as highly parallel performance and therefore significant enhancement in speed can be obtained. High accurate direction estimate of incoming signals is then derived through the proposed algorithm with the technique of norm-1 followed by interpolation. Finally computer simulation is given to indicate that the proposed algorithm has met the challenge and ensures high accuracy. Compared to its CPU based algorithm, better real-time performance can be obtained on its GPU version.

总装预研基金资助项目;中央高校基本科研项目; 国家自然科学基金资助项目