基于GPU的Landsat8实时解压缩处理技术

根据Landsat8卫星下行数据的特点,结合RICE无损数据压缩算法和CUDA编程技术,提出一种基于图像处理单元（GPU）的实时解压缩处理系统方案,采用统一计算设备架构和流水线式解压缩结构进行任务分解及CUDA优化。测试结果表明,与基于CPU的解压缩方案相比,基于GPU的解压缩系统获得7.45倍的加速比,能够达到Landsat8卫星下行数据实时解压缩处理速度要求的1.8倍以上,解压缩的图像数据正确率高且成本较低。...

Full description

Saved in:
Bibliographic Details
Published in计算机工程 Vol. 42; no. 3; pp. 301 - 307
Main Author 杨仁忠 张洁 韦宏卫 石璐
Format Journal Article
LanguageChinese
Published 中国科学院遥感与数字地球研究所,北京,100094%中国科学院遥感与数字地球研究所,北京100094 2016
中国科学院大学,北京100049
Subjects
卫星下行数据
图形处理单元
实时解压缩
并行计算
无损数据压缩算法
lossless data compression algorithm
图形处理单元
real-time decompression
satellite downlink data
卫星下行数据
并行计算
Graphic Processing Unit (GPU)
parallel computing
无损数据压缩算法
实时解压缩
Online AccessGet full text
ISSN1000-3428
DOI10.3969/j.issn.1000-3428.2016.03.054

Cover

More Information
Summary:根据Landsat8卫星下行数据的特点,结合RICE无损数据压缩算法和CUDA编程技术,提出一种基于图像处理单元（GPU）的实时解压缩处理系统方案,采用统一计算设备架构和流水线式解压缩结构进行任务分解及CUDA优化。测试结果表明,与基于CPU的解压缩方案相比,基于GPU的解压缩系统获得7.45倍的加速比,能够达到Landsat8卫星下行数据实时解压缩处理速度要求的1.8倍以上,解压缩的图像数据正确率高且成本较低。
Bibliography:31-1289/TP
YANG Renzhong;ZHANG Jie;WEI Hongwei;SHI Lu; （1. Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences ,Beijing 100094 ,China; 2. University of Chinese Academy of Sciences, Beijing 100049,China）
According to the features of downlink data of Landsat8 as well as the RICE lossless data compression algorithm and the CUDA programming technology,a real-time decompression processing system bsed on Graphic Processing Unit（ GPU） is proposed. It adopts Compute Unified Device Architecture（ CUDA） and pipelined decompression structure for task decompression and CUDA optimization. Experimental results showthat the speed-up of the implemented decompression system is 7. 45 times that of the CPU-based solution,and the processing rate is at least1. 8 times as high as that required by Landsat8 satellite downlink data real-time decompression. The decompressed image data has high accuracy with lowcost.
lossless data compression algorithm; real-time decompression; Graphic Processing Unit（GPU）; satellite
ISSN:1000-3428
DOI:10.3969/j.issn.1000-3428.2016.03.054