计算机科学 ›› 2020, Vol. 47 ›› Issue (4): 25-29.doi: 10.11896/jsjkx.190500029

• 计算机体系结构 • 上一篇    下一篇

基于GPU多流并发并行模型的NDVI提取算法

左宪禹1,2, 张哲1,3, 苏岳瀚1, 刘扬1,2, 葛强1,2, 田军锋1,2   

  1. 1 河南大学计算机与信息工程学院数据与知识工程研究所 河南 开封475004;
    2 河南省大数据分析与处理重点实验室 河南 开封475004;
    3 中国科学院空天信息创新研究院 北京100094
  • 收稿日期:2019-05-06 出版日期:2020-04-15 发布日期:2020-04-15
  • 通讯作者: 田军锋(tjf328@126.com)
  • 基金资助:
    国家重点研发计划课题(2017YFD0301105);国家自然科学基金(U1704122,U1604145);河南省重点研发与推广专项(182102210242,182102110065,192102210096)

Extraction Algorithm of NDVI Based on GPU Multi-stream Parallel Model

ZUO Xian-yu1,2, ZHANG Zhe1,3, SU Yue-han1, LIU Yang1,2, GE Qiang1,2, TIAN Jun-feng1,2   

  1. 1 Institute of Data and Knowledge Engineering,College of Computer and Information Engineering,Henan University,Kaifeng,Henan 475004,China;
    2 Henan Key Laboratory of Big Data Analysis and Processing,Henan University,Kaifeng,Henan 475004,China;
    3 Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China
  • Received:2019-05-06 Online:2020-04-15 Published:2020-04-15
  • Contact: TIAN Jun-feng,born in 1980,Ph.D,lecturer.His main research interests include remote sensing image proces-sing and image encryption.
  • About author:ZUO Xian-yu,born in 1979,Ph.D,associate professor.His research interests include parallel computing and remote sensing image processing.
  • Supported by:
    This work was supported by the National Key Research and Development Program (2017YFD0301105),National Natural Science Foundation of China(U1704122,U1604145) and Key R&D and Promotion Projects of Henan Province (182102210242,182102110065,192102210096).

摘要: 利用GPU进行加速的归一化差分植被指数(Normalized Differential Vegetation Index,NDVI)提取算法通常采用GPU多线程并行模型,存在弱相关计算之间以及CPU与GPU之间数据传输耗时较多等问题,影响了加速效果的进一步提升。针对上述问题,根据NDVI提取算法的特性,文中提出了一种基于GPU多流并发并行模型的NDVI提取算法。通过CUDA流和Hyper-Q特性,GPU多流并发并行模型可以使数据传输与弱相关计算、弱相关计算与弱相关计算之间达到重叠,从而进一步提高算法并行度及GPU资源利用率。文中首先通过GPU多线程并行模型对NDVI提取算法进行优化,并对优化后的计算过程进行分解,找出包含数据传输及弱相关性计算的部分;其次,对数据传输和弱相关计算部分进行重构,并利用GPU多流并发并行模型进行优化,使弱相关计算之间、弱相关计算和数据传输之间达到重叠的效果;最后,以高分一号卫星拍摄的遥感影像作为实验数据,对两种基于GPU实现的NDVI提取算法进行实验验证。实验结果表明,与传统基于GPU多线程并行模型的NDVI提取算法相比,所提算法在影像大于12000*12000像素时平均取得了约1.5倍的加速,与串行提取算法相比取得了约260倍的加速,具有更好的加速效果和并行性。

关键词: GPU多流并发模型, NDVI, 并行加速, 计算通讯重叠, 遥感信息提取

Abstract: In general,the Normalized Differential Vegetation Index (NDVI) extraction algorithm optimized by GPU usually adopts GPU multi-thread parallel model,and there are problems such as data transmission between CPU and GPU and weak correlation calculations taking more time,which affect the further improvement of performance.Aiming at the above problems and the characteristics of NDVI,a NDVI extraction algorithm based on GPU multi-stream parallel model was proposed.Through the features of CUDA stream and Hyper-Q,the GPU multi-stream parallel model can overlap not only data transmission and kernel execution,but also kernel execution and kernel execution,and further improve parallelism and resources utilization of GPU.Firstly,the NDVI algorithm is optimized by the GPU multi-thread parallel model,and the optimized procedures are decomposed to find out the parts of the algorithm with data transmission or weak correlation calculation.Secondly,parts of data transmission and weak correlation calculation are reconstructed and optimized by GPU multi-stream parallel model to achieve overlapping between weak correlation calculation and weak correlation calculation,or weak correlation calculation and data transmission.Finally,expe-riments of NDVI algorithm that based on both GPU parallel models respectively were carried out,and the remote sensing image taken by the GF1 satellite were used as experimental data.The experimental results show that the proposed algorithm,when the image is larger than 12000x13400 pixels,achieves about 1.5 times acceleration compared with the traditional parallel NDVI algorithm based on the GPU multi-thread parallel model,and about 260 times acceleration compared with the NDVI sequential extraction algorithm,which has better performance and parallelism.

Key words: GPU parallel model, NDVI, Overlap, Parallel acceleration, Remote sensing information extraction

中图分类号: 

  • TP751
[1]ZUO X Y,SHANG D D,LI B B,et al.Parallel Computing Reasearch of Normalized Difference Vegetation Index Based on OpenMP and OpenCV[J].Remote Sensing Science,2017,5:33-40.
[2]MENG H.Research on remote sensing image normalized difference vegetation index based on GPU[D].Kaifeng:HeNan University,2016.
[3]ALVAREZ C J,HERRERA L J,RIVERA Z I,et al.Implementation Strategy of NDVI Algorithm with Nvidia Thrust [C]//6th Pacific-Rim Symposium on Image and Video Technology.2013:184-193.
[4]JI X,LU H Y,ZHAO T J,et al.Study on climate mean based on MODIS vegetation index[J].Journal of Guangxi University(Nat SciEd),2018,43(3):1111-1117.
[5]DI P,HU C J,LI J J,et al.Efficient method for histogram generation on GPU[J].Computer Science,2012,39(3):304-307.
[6]AHAMD L,EHSAN A,AMIRALI B,et al.TELEPORT:Hardware/software alternative to CUDA shared memory progra-mming[J].Micro processors and Microsystems,2018,63:169-181.
[7]CAO J,HUANG K J,WANG J H,et al.Particle filter multispeakers tracking algorithm based on GPU and its application[J].Application Research of Computers,2018,35(7):1965-1969.
[8]SHEN X J,HOU B C,HAN D J,et al.Calculation of enhanced vegetation index based on GPU and matrix partition[J].Remote Sensing Information,2018,33(3):63-69.
[9]GRUBOV V,MAKSIMENKO V A,NEDAI-VOZOV V,et al.Real-Time Big EEG Data Processing With CUDA Parallel Computing Technology [C]//2018 2nd School on Dynamics of Complex Networks and their Application in Intellectual Robotics (DCNAIR).2018:49-52.
[10]KEI I,FUMIHIKO I,KENICHI H,et al.An OpenACC Optimizer for Accelerating Histogram Computation on a GPU[C]//Euromicro International Conference on Parallel,Distributed,and Network-Based Processing.2016:468-477.
[11]WU Q J,CHEN Y M,JOHN P W,et al.An effective parallelization algorithm for DEM generalization based on CUDA[J].Environmental Modelling & Software,2019,114:64-74.
[12]LIU Z T,YAN B X,DONG M L,et al.Application of parallel computing in edge extraction algorithm in dynamic photogrammetry[J].Computer Engineering and Design,2019,40(1):97-102 .
[13]HONG H C,ZHENG L X,PAN S W,et al.Computation of Gray Level Co-Occurrence Matrix Based on CUDA and Optimization for Medical Computer Vision Application [J].IEEE Access,2018,6:67762-67770.
[14]RENAN P,CRISTIANA B,RICARDO F,et al.Video Processing on GPU:Analysis of Data Transfer Overhead[C]//IEEE International Symposium on Computer Architecture and High Performance Computing Workshops.2016:18-23.
[15]MA X,HAN W,et al.A Parallel Multi-swarm Particle Swarm Optimization Algorithm Based on CUDA Streams[C]//2018 Chinese Automation Congress (CAC).2018:3002-3007.
[16]LI H,YU D,KUMAR A,et al.Performance modeling in CUDA streams — A means for high-throughput data processing[C]//IEEE International Conference on Big Data (Big Data).2014:301-310.
[17]RYAN S L,QIU Q R,et al.Effective Utilization of CUDA Hyper-Q for Improved Power and Performance Efficiency[C]//IEEE International Parallel and Distributed Processing Sympo-sium Workshops.2016:1160-1169.
[18]SREEPATHI P,THAZHUTHAVEETIL M J,GOVINDARAJAN R,et al.Improving GPGPU concurrency with elastic kernels[C]//ASPLOS’13.ACM,2013:407-418.
[1] 汪亮, 周新志, 严华.
基于GPU的实时SIFT算法
Real-time SIFT Algorithm Based on GPU
计算机科学, 2020, 47(8): 105-111. https://doi.org/10.11896/jsjkx.190700036
[2] 洪朝群,陈旭辉,王晓栋,李士锦,吴克寿.
基于GPU并行加速的多特征融合的超图降维方法
Hypergraph Dimensionality Reduction with Multiple Feature Fusion Based on GPU Parallel Acceleration
计算机科学, 2015, 42(11): 90-93. https://doi.org/10.11896/j.issn.1002-137X.2015.11.018
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!