CUDA并行加速的稀疏PCNN运动目标检测算法  

Processing for accelerated sparse PCNN moving target detection algorithm with CUDA

作　　者：凌滨[1] 邓艳[1] 于士博[1] 

机构地区：[1]东北林业大学机电工程学院,黑龙江哈尔滨150040

出　　处：《计算机工程与设计》2016年第12期3300-3305,3315,共7页Computer Engineering and Design

摘　　要：为准确检测低速径向运动的小运动目标,降低系统的噪声,提高系统的实时性,提出一种基于Nvidia通用并行计算架构(CUDA)的稀疏脉冲耦合神经网络运动目标检测的并行算法。根据图形处理单元(GPU)的并行结构和硬件特点,将改进帧差法得到二值图像的过程,以及差分二值图像映射到稀疏脉冲耦合神经网络模型的过程均放GPU上执行,提高算法的计算效率;选择利用纹理存储和共享存储方式,提高数据的访问效率,降低算法的复杂度。实验结果表明,该算法对运动目标检测的准确性和实时性优于其它方法。To detect the small moving targets with low velocity and radial motion,to reduce the system noise and improve the system real-time performance,aparallel algorithm for moving object detection in sparse pulse coupled neural network based on Nvidia compute unified device architecture（CUDA）was proposed.According to the parallel architecture and hardware characteristics of the graphics processing unit（GPU）,the improved frame difference method was used in the process of getting the two value image,and the differential two value image was mapped to the sparse pulse coupled neural network model,which was put on GPU to improve the computational efficiency of the algorithm.At the same time,to improve the efficiency of data access and reduce the complexity of the algorithm,the use of texture memory and shared memory was selected.Experimental results were compared with other methods,and the accuracy and real-time performance of the proposed algorithm are verified.

关 键 词：通用并行计算架构 稀疏脉冲耦合神经网络 改进帧差法 运动目标检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]