一种基于 MPI 和 OpenMP 的剖分遥感影像并行分割方法

摘要

随着遥感数据量的快速增长，组织效率和处理速率成为其快捷应用的瓶颈，地球剖分理论和高性能计算为该问题解决提供了一种可能途径。基于上述思路，设计一种基于 MPI 和 OpenMP 的剖分遥感影像并行分割方法。该方法在地球剖分组织理论的研究基础上，针对遥感影像数据进行剖分化处理，依据应用需求确定剖分层级，形成有利于并行处理的影像数据剖分面片集合；在MPI 集群计算和 OpenMP 多核任务处理环境下，对上述剖分面片集进行相应操作的并行化处理，最后把处理后的面片整合为所需要的结果数据。通过一个遥感影像数据 K-Means 分割的例子对上述方法进行验证。结果表明，该方法能在分割结果保持一致的情况下，有效提高遥感影像的组织效率和处理速度，达到了一定的加速比，实用性较强。%With the rapid growth of remote sensing data,organisational efficiency and processing speed become the bottlenecks of its fast applications,but earth partition theory and high-performance computing technology offer a possible solution.Guided by this idea,the researchers designed an MPI-and-OpenMP-based parallel segmentation method for partitioned remote sensing images.On the basis of earth partition organisation theory study,the method conducts partition process aimed at remote sensing image data and determines the partition layer according to the application requirements,and forms the partitioning facets set of image data conducive to parallel processing;Under the context of MPI cluster computing and OpenMP multi-core tasks processing,the researchers carried out the parallel processing of corresponding operation on above mentioned partitioning facets set,and finally integrated the processed facets into required result data.The above method has been verified by an example of K-means segmentation of remote sensing image data.Experimental result indicated that under the condition of keeping the conformity of segmentation results,this method could effectively raise the organisational efficiency and processing speed of remote sensing image,reached certain speedup and had strong practicality.