计算机与现代化

• 软件工程 • 上一篇    下一篇

基于MPI+OpenMP的红外弱小目标检测并行计算

  

  1. (国防科学技术大学电子科学与工程学院,湖南长沙 410072)
  • 收稿日期:2014-05-08 出版日期:2014-07-16 发布日期:2014-07-17
  • 作者简介:贺维维(1990-),女,湖北随州人,国防科学技术大学电子科学与工程学院硕士研究生,研究方向:目标检测与识别,并行计算; 吴京(1965-),女,教授,硕士生导师,博士,研究方向:数字信号处理; 曾瑶源(1986-),男,讲师,博士,研究方向:高性能计算。
  • 基金资助:
    中国博士后科学基金资助项目(2013M532167)

Parallel Computing of Infrared Small Target Detection Based on MPI and OpenMP

  1. (College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410072, China)
  • Received:2014-05-08 Online:2014-07-16 Published:2014-07-17

摘要: 为有效监控红外弱小目标运动的全过程,必须采用多个波段同时探测,但是多波段探测必然带来计算时间的大幅增长,无法满足实际应用中对目标检测实时性的要求。针对这一问题,本文提出一种基于MPI+OpenMP的层次化并行方法,充分利用消息传递模型和共享存储模型的优势,并基于多处理器节点集群进行测试。实验结果表明,该并行程序在保证相同的检测概率的情况下加速比达到8.61,极大地提高了目标检测的效率。

关键词: 目标检测, MPI, OpenMP, 并行计算

Abstract: To get effective surveillance of the whole movement of infrared small target, multiband simultaneous detection is necessary. But, multiband simultaneous detection leads to substantial growth in computing time, that cannot meet the requirement of real-time target detection in practical application. To solve this problem, this paper presents a parallel hierarchical approach based on MPI and OpenMP, which makes full use of the advantages of both message-passing model and shared storage model. It is based on multi-processor node cluster for testing. Experimental results show that, this parallel version of the program can achieve a speedup of 8.61 compared to the serial version in the condition of same detection probability. It improves the efficiency of target detection greatly.

Key words: target detection, MPI, OpenMP, parallel computing

中图分类号: