基于MPI、MapReduce和OpenMP混合编程的高分三号数据分布式并行转换算法
Distributed Parallel Conversion Algorithm of GF-3 Data Based on MPI,MapReduce and OpenMP Hybrid Programming
史超 1蔡源浩 2陈超 1姚术林 2陈云2
作者信息
- 1. 呼伦贝尔市大数据中心
- 2. 呼伦贝尔智慧城市发展研究院 内蒙古呼伦贝尔 021000
- 折叠
摘要
高分三号是我国C波段多极化合成孔径雷达卫星,形成三星组网星座.PolSARpro是欧空局支持下的一款SAR影像处理的开源软件.为了便于使用该软件来处理高分三号数据,提出了一种根据PolSARpro软件的数据格式要求进行分布式并行转换算法.该算法内层采用OpenMP按极化方式并行计算,中间层采用MapReduce按景并行处理,外层采用MPI按任务分布式计算,最终实现将高分三号极化数据快速精确转化为复数散射矩阵S2数据格式.通过KingMap V10.0平台实现了算法并在实际数据中进行测试,验证了算法的高效性、可行性和正确性.
Abstract
Gaofen-3 is the C-band multi-polarization SAR satellite in China,and networked a constellation of 3 satellites.PolSARpro is an open source software for SAR image processing supported by ESA.In order to process GF-3 data conveniently,this paper proposes a distributed parallel conversion algorithm according to the data format requirements of its software,internal layer based on OpenMP by polarization,intermediate layer based on MapRe-duce by scene,and external layer based on MPI by task,which can quickly and accurately convert GF-3 polariza-tion data into complex sinclair matrix data format.The algorithm is implemented on KingMap V10.0 platform and tested in real data,which verifies the efficiency,feasibility and correctness of the algorithm.
关键词
高分三号/复数散射矩阵/合成孔径雷达/并行计算/分布式计算Key words
GF-3/Complex sinclair matrix/Synthetic aperture radar/Parallel computing/Distributed computing引用本文复制引用
出版年
2024
NETL
NSTL
万方数据