文献类型：期刊文献

中文题名：基于MPI的LEDAPS遥感影像预处理并行化方法研究

英文题名：LEDAPS remote sensing image processing parallelization method research based on MPI

作者：陈雄华[1] 张旭[1] 郭颖[1] 马勇[1] 杨彦臣[1]

第一作者：陈雄华

机构：[1]中国林业科学研究院资源信息研究所

年份：2013

卷号：40

期号：11

起止页码：201-205

中文期刊名：广东农业科学

外文期刊名：Guangdong Agricultural Sciences

收录：CSTPCD;;北大核心:【北大核心2011】；CSCD:【CSCD_E2013_2014】；

基金：中央级公益性科研院所基本科研业务费专项(IFRIT201104)

语种：中文

中文关键词：森林碳储量;LEDAPS;Landsat影像预处理;MPI;并行计算

外文关键词：forest carbon stocks; LEDAPS; Landsat image preprocessing; MPI; parallel computing

分类号：S771.8

摘要：LEDAPS(Landsat生态系统干扰自适应处理系统)通过对Landsat影像进行定标、云掩模、精确配准和正射纠正、大气校正等预处理,为森林生态系统固碳能力及碳储量研究提供地表反射率产品。随着遥感数据的几何增长,传统串行使用LEDAPS进行影像预处理计算所费周期长,使得LEDAPS在实际森林碳储量研究应用中不能满足海量遥感数据的处理需求。针对这一问题,提出了一种基于MPI的LEDAPS高性能粗粒度数据并行计算方法。通过实例验证,当MPI进程数为8时,加速比最高达到7.37。该方法在大幅提高计算速度,节省计算时间的基础上,实现了计算节点的负载均衡及可扩展,有效地提高了LEDAPS处理海量遥感数据的能力,缩短了利用遥感影像进行森林碳储量计算的周期。
Based on Landsat image, the Landsat Ecosystem Disturbance Adaptive Processing System （LEDAPS） uses radiation change detection method for image processing and offers the surface reflectivity products for ecosystem carbon sequestration and carbon reserves. As the accumulation of massive remote sensing data,the traditional serial LEDAPS for image processing has a long cycle that make a lot of difficulties in practical application. For this problem, this paper design a high performance parallel computing method based on MPI. Research and experiment show that the highest speed ratio reached 7.37 when the number of MPI process is 8. The method not only greatly improve the calculation speed and save computing time, but also realize the load balance between the computing nodes and the computing nodes can be extended. It effectively improves the ability of LEDAPS to handle massive remote sensing data and reduces the forest carbon stocks calculation cycle by using the remote sensing images.