基于GPU并行分块的二维恒虚警检测算法的实现方法

Implementation method of two-dimensional constant false alarm rate detection algorithm based on GPU parallel block

相对于一维恒虚警检测,二维恒虚警检测具有更优的杂波边缘和多目标检测性能。但二维恒虚警检测方法需要进行距离-多普勒两维计算,导致算法难以实时实现。针对这些问题,在基于图形处理器(GPU)和中央处理器(CPU)的架构下,提出了并行分块存储技术和块间数据并行归约求和方法,并结合统一计算设备架构(CUDA)技术,实现基于GPU的二维恒虚警检测算法并行处理。实验结果表明,相比于传统的CPU实现,所提的实现方法不仅具有优异杂波边缘检测性能,并且加速比达到了600倍以上,此外随着计算量的增加,加速效果更为显著,能够满足系统实时性和大吞吐量的要求。

Compared with one-dimensional constant false alarm rate detection,two-dimensional constant false alarm rate detection has better clutter edge and multi-target detection performance.However,the two-dimensional constant false alarm rate detection method requires distance-Doppler two-dimensional calculation,which makes the algorithm difficult to implement in real-time.To address this issue,this paper proposes a parallel block storage technology and a parallel scan reduction summation method based on the hardware structure characteristics of GPU.The implementation method of two-dimensional constant false alarm rate detection algorithm based on GPU parallel block is realized by using CUDA technology under the architecture of GPU and CPU.Experimental results indicate that compared to the traditional CPU implementation,the implementation method proposed in this paper not only has excellent clutter edge detection performance,but also achieves a speedup of more than 600 times.Furthermore,with the increase in computation,the acceleration effect is more significant,able to meet the requirements of real-time and high-throughput systems.