遥感影像数据并行计算中数据分配策略研究  

Research on data distribution strategy in parallel computing of remote sensing image

作　　者：马伟锋[1,2] 李伟[2] 

机构地区：[1]浙江科技学院信息与电子工程学院,浙江杭州310023 [2]浙江工业大学浙江省可视媒体智能处理技术研究重点实验室,浙江杭州310023

出　　处：《浙江工业大学学报》2016年第3期270-274,共5页Journal of Zhejiang University of Technology

基　　金：浙江省可视媒体智能处理技术研究重点实验室开放基金资助项目(2012007)

摘　　要：分布式并行计算技术的发展使得遥感影像的快速处理成为了可能,而数据分配策略是遥感影像数据并行计算的核心问题.针对不同的影像数据类型和计算算法,提出了基于计算优化的平均划分与非平均划分和基于效果增强的窗口划分与尺度划分四种数据分配策略模式,分析了不同模式适合的算法类型和应用场合,对基于计算优化的模式进行计算速度和效率的理论分析,并通过构建MPI并行计算环境进行实验测试和分析.实验结果符合理论分析预期,可以对遥感影像数据并行计算系统构建和算法设计提供有效的参考和评价.The development of distributed and parallel computing technology enables rapid processing of remote sensing image possible,and data distribution strategy is the core issue in remote sensing image parallel computing.This paper analyzes the current remote sensing parallel computing and data distribution methods.Base on the current status of research work,four data distribution strategy modes are proposed such as evenly division,non-evenly division,windowbased division and scale division based on calculation optimization.Moreover,the paper discusses that the four distribution strategy modes would adapt to different types of algorithms and applications.The computing speed and efficiency are analyzed theoretically for the four modes based on calculation optimization. Finally, through building MPI parallel computing environment,the four data distribution strategies are tested and analyzed The experimental results correspond to the expected theoretical analysis results.It will be useful for system construction and algorithm design of the remote sensing image parallel computing.

关 键 词：遥感计算 并行计算 数据分配 遥感影像处理 分布式计算 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]