Authors:
Shamim Akhter
1
;
Kento Aida
2
and
Yann Chemin
3
Affiliations:
1
Tokyo Institute of Technology, Japan
;
2
Tokyo Institute of Technology; National Institute of Informatics, Japan
;
3
International Rice Research Institute (IRRI), Philippines
Keyword(s):
NDVI, Remote Sensing, LMF, High Performance Computing, Grid.
Related
Ontology
Subjects/Areas/Topics:
Agents
;
Artificial Intelligence
;
Distributed and Mobile Software Systems
;
e-Business
;
Energy and Economy
;
Energy-Aware Systems and Technologies
;
Enterprise Information Systems
;
Geographic Information Systems (GIS)
;
Grid Computing
;
Internet Technology
;
Parallel and High Performance Computing
;
Software Engineering
;
Technology Platforms
;
Web Information Systems and Technologies
Abstract:
Vegetation Index Map provides the crop density information over a precise region. Remote Sensing (RS) images are at the basis of creating such map, while the decision-maker requirement stands for Vegetation Index Maps at various in-country administrative levels. However, RS image includes data noises due to influence of haze or cloud especially in the rainy season. Temporally Splined procedure such as Local Maximum Fitting (LMF) can be applied on RS images for ensuring the data consistency. Running the LMF procedure with single computer takes impractical amount of processing time (approx. 150 days) for Asia regional RS image (46 bands/dates, 3932 rows, 11652 columns). Importing the LMF on High Performance Computing (HPC) platforms provides with a time optimization mechanism, and LMF has been implemented in cluster computers for this very purpose. A single cluster LMF processing timing still did not perform within an acceptable time range. In this paper, the LMF processing methodology
to reduce processing time by combining the parallelization of data and task together on multi-cluster Grids is presented.
(More)