PARALLEL SPARSE UNMIXING OF HYPERSPECTRAL DATA

被引：9

作者：

Rodriguez Alves, Jose M. ^{[1
]}

Nascimento, Jose M. P. ^{[1
,2
]}

Bioucas-Dias, Jose M. ^{[1
,3
]}

Plaza, Antonio ^{[4
]}

Silva, Vitor ^{[5
]}

机构：

[1] Inst Telecomunicacoes, Lisbon, Portugal

[2] Inst Super Engn Lisboa, Lisbon, Portugal

[3] Univ Tech Lisbon, Inst Super Tecn, P-1100 Lisbon, Portugal

[4] Univ Extremadura, Hyperspectral Comp Lab, Caceres, Spain

[5] Univ Coimbra, Inst Telecommunicacoes, DEEC, P-3000 Coimbra, Portugal

来源：

2013 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM (IGARSS) | 2013年

关键词：

Hyperspectral Unmixing; Sparse Regression; Graphics Processing Unit; Parallel Methods; ENDMEMBER EXTRACTION; COMPONENT ANALYSIS;

D O I：

10.1109/IGARSS.2013.6723057

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

In this paper, a new parallel method for sparse spectral unmixing of remotely sensed hyperspectral data on commodity graphics processing units (GPUs) is presented. A semi-supervised approach is adopted, which relies on the increasing availability of spectral libraries of materials measured on the ground instead of resorting to endmember extraction methods. This method is based on the spectral unmixing by splitting and augmented Lagrangian (SUNSAL) that estimates the material's abundance fractions. The parallel method is performed in a pixel-by-pixel fashion and its implementation properly exploits the GPU architecture at low level, thus taking full advantage of the computational power of GPUs. Experimental results obtained for simulated and real hyperspectral datasets reveal significant speedup factors, up to 1 64 times, with regards to optimized serial implementation.

引用

页码：1446 / 1449

页数：4