基于Matlab的高光谱遥感数据降维并行计算分析

Parallel Computation in Dimension Reduction of Hyperspectral Remote Sensing Imagery Based on Matlab

对于海量遥感数据的计算而言,串行运算对计算机性能要求高,而且耗时长。为此本文提出引用并行运算方法,不仅可以降低对计算机性能的要求,还可以大大提高运行和计算速度。为此,首先介绍了基于MPI(Message Passing Interface)的并行运算机制,且以Matlab为例给出了它的并行模式,并详细介绍了将现有串行运算代码改造成并行运算的流程。以海量高光谱影像数据为例,将本征维数估计的串行运算修改为并行运算,实验分析并测试了其运行效率。结果表明,并行计算较串行计算可大大缩短本征维数的计算时间。

Serial computation usually requires computer of high-performance and time-consuming when calculation of huge amount of remote sensing data.So the parallel computation is involved in to avoid the strict request of computer capacity,and help to improve the speed of calculation significantly.The mechanism of parallel computation based on MPI（Message Passing Interface） is introduced and performed a model in MATLAB as well.The detail flowchart of the way how to change the serial computation code into the parallel one is then presented.As a case study,the code of intrinsic dimension estimation of large volume hyperspectral remote sensing data was changed from serial computation mode into parallel computation.Finally,numerical results show that parallel computation has clearly advantage in efficiency when perform the complicated computation.