基于MPI和OpenCV遥感植被指数产品的并行计算

遥感传感器和计算机技术的发展，每天都会汇集大量新的地理空间数据。地球科学许多应用要求数据实时或接近实时地处理，发展高性能计算是进行海量数据处理的必然趋势。本文以 TM 影像制备黑河流域归一化指数产品为例，基于高性能集群，实现了植被指数快速提取的并行计算方法，并采用对等并行编程模式，通过 C 语言调用 MPI（Message Passing Interface，消息传递接口）和 OpenCV（Open Source Computer Vision Library，开源计算机视觉库）函数库，实现了 NDVI（Normalized Difference Vegetation Index，归一化植被指数）的并行计算，获得了黑河流域的 NDVI。性能测试表明，并行计算可以显著提高遥感图像处理的速度。文章最后讨论了从原始影像提取植被指数产品的流程。

Parallel Computing of Vegetation Index Products Based on MPI and OpenCV

With the advances of sensors and computer technology, a large number of new geospatial data is collected every day. In particular, many current and future applications of remote sensing in earth science require real or near real-time processing capabilities. Employing high performance computing (HPC) for remote sensing missions is an inevitable trend. As an example, we have made products of normalized difference vegetation index in Heihe watershed, and this paper presents how to implement parallel processing of remote sensing image based on high performance cluster. We use C language combined with MPI and OpenCV library and apply peer-to-peer parallel programming model to realize the parallel computation of NDVI. The performance test shows that this parallel method can significantly improve the speed of remote sensing image processing. At the end, the procedure of making vegetation index products from the original image is discussed.