InSAR图的计算统一构架干涉滤波并行计算
|
摘要
作为InSAR处理中的重要操作,干涉图滤波是比较耗时的步骤之一。针对广泛应用的Goldstein滤波算法,文章设计了一种基于GPU的滤波加速方法,定量评价了滤波效果,并探讨了基于GPU加速的滤波时间与数据大小和滤波窗口大小的关系,讨论了当前GPU的核心设计在并行计算中的缺点。在取得与CPU处理结果相同精度的前提下,此方法获得大约22倍的加速比。
As an important step in InSAR processing,interferogram filtering is one of the most timeconsuming step.In the paper,a method of filtering acceleration based on GPU aiming at the widely used Goldstein filtering algorithm was proposed,the filtering effects were quantitatively evaluated,and the relationship between the filtering time with GPU and the data volumes and the filtering kernel sizes were discussed then.Furthermore,the drawbacks of current GPUs in parallel computing were pointed out.Experimental result proved that the method could achieve the same accuracy as CPUs do,with a speedup ratio as high as 22.
As an important step in InSAR processing,interferogram filtering is one of the most timeconsuming step.In the paper,a method of filtering acceleration based on GPU aiming at the widely used Goldstein filtering algorithm was proposed,the filtering effects were quantitatively evaluated,and the relationship between the filtering time with GPU and the data volumes and the filtering kernel sizes were discussed then.Furthermore,the drawbacks of current GPUs in parallel computing were pointed out.Experimental result proved that the method could achieve the same accuracy as CPUs do,with a speedup ratio as high as 22.
引文
[1]卢丽君.InSAR影像配准及其并行化算法研究[D].武汉:武汉大学,2005.
[2]李颖.基于MPI的InSAR两个关键技术的并行化设计与实现[D].北京:北京大学,2008.
[3]Giancaspro A,Candela L,Lopinto E,et al.SAR images co-registration parallel implementation[C]//Proc.on Geoscience and Remote Sensing Symposium.Toronto:IGARSS'02,2002:1337-1339.
[4]王萍,汪长城,彭星,等.基于PolInSAR三分量分解的建筑物高度向信息提取方法[J].测绘工程,2014,23(6):16-20.
[5]温婵娟,欧嘉蔚,贾金原.GPU通用计算平台上的SPH流体模拟[J].计算机辅助设计与图形学学报,2010,22(3):406-411.
[6]杨靖宇,张永生,董广军.基于GPU的遥感影像SAM分类算法并行化研究[J].测绘科学,2010,35(3):9-11.
[7]刘洲俊,胡包钢.GPU加速的高分辨率DEM图像地形特征线提取算法[J].中国图象图形学报,2012,17(2):249-255.
[8]Zebker HA,Villasenor J.Decorrelation in interferometric radar echoes[C]//IEEE Transactions on Geoscience and Remote Sensing,1992.30(5):950-959.
[9]Goldstein RM,Werner CL.Radar interferogram filtering for geophysical applications[J].Geophysical Research Letters,1998,25(21):4035-4038.
[10]NVIDIA.CUDA Toolkit 4.1 CUFFT Library[EB/OL].(2012).http://www.metz.supelec.fr/metz/personnel/vialle/course/Mineure-HPC/doc-cuda/CUFFT_Library.pdf.
[11]曾琪明.合成孔径雷达干涉技术研究与应用——以中国台湾集集921地震为例[D].北京:北京大学,2000.
[2]李颖.基于MPI的InSAR两个关键技术的并行化设计与实现[D].北京:北京大学,2008.
[3]Giancaspro A,Candela L,Lopinto E,et al.SAR images co-registration parallel implementation[C]//Proc.on Geoscience and Remote Sensing Symposium.Toronto:IGARSS'02,2002:1337-1339.
[4]王萍,汪长城,彭星,等.基于PolInSAR三分量分解的建筑物高度向信息提取方法[J].测绘工程,2014,23(6):16-20.
[5]温婵娟,欧嘉蔚,贾金原.GPU通用计算平台上的SPH流体模拟[J].计算机辅助设计与图形学学报,2010,22(3):406-411.
[6]杨靖宇,张永生,董广军.基于GPU的遥感影像SAM分类算法并行化研究[J].测绘科学,2010,35(3):9-11.
[7]刘洲俊,胡包钢.GPU加速的高分辨率DEM图像地形特征线提取算法[J].中国图象图形学报,2012,17(2):249-255.
[8]Zebker HA,Villasenor J.Decorrelation in interferometric radar echoes[C]//IEEE Transactions on Geoscience and Remote Sensing,1992.30(5):950-959.
[9]Goldstein RM,Werner CL.Radar interferogram filtering for geophysical applications[J].Geophysical Research Letters,1998,25(21):4035-4038.
[10]NVIDIA.CUDA Toolkit 4.1 CUFFT Library[EB/OL].(2012).http://www.metz.supelec.fr/metz/personnel/vialle/course/Mineure-HPC/doc-cuda/CUFFT_Library.pdf.
[11]曾琪明.合成孔径雷达干涉技术研究与应用——以中国台湾集集921地震为例[D].北京:北京大学,2000.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心