Please use this identifier to cite or link to this item: http://hdl.handle.net/10553/35376
Title: SVM-based real-time hyperspectral image classifier on a manycore architecture
Authors: Madroñal, D.
Lazcano, R.
Salvador, R.
Fabelo, H. 
Ortega, S. 
Callico, G. M. 
Juarez, E.
Sanz, C.
UNESCO Clasification: 33 Ciencias tecnológicas
Keywords: Support Vector Machine
Hyperspectral imaging
Massively parallel processing
Real-time processing
Energy consumption awareness, et al
Issue Date: 2017
Journal: Journal of Systems Architecture 
Conference: Conference on Design and Architectures for Signal and Image Processing (DASIP) 
Abstract: This paper presents a study of the design space of a Support Vector Machine (SVM) classifier with a linear kernel running on a manycore MPPA (Massively Parallel Processor Array) platform. This architecture gathers 256 cores distributed in 16 clusters working in parallel. This study aims at implementing a real-time hyperspectral SVM classifier, where real-time is defined as the time required to capture a hyperspectral image. To do so, two aspects of the SVM classifier have been analyzed: the classification algorithm and the system parallelization. On the one hand, concerning the classification algorithm, first, the classification model has been optimized to fit into the MPPA structure and, secondly, a probability estimation stage has been included to refine the classification results. On the other hand, the system parallelization has been divided into two levels: first, the parallelism of the classification has been exploited taking advantage of the pixel-wise classification methodology supported by the SVM algorithm and, secondly, a double-buffer communication procedure has been implemented to parallelize the image transmission and the cluster classification stages. Experimenting with medical images, an average speedup of 9 has been obtained using a single-cluster and double-buffer implementation with 16 cores working in parallel. As a result, a system whose processing time linearly grows with the number of pixels composing the scene has been implemented. Specifically, only 3 mu s are required to process each pixel within the captured scene independently from the spatial resolution of the image.
URI: http://hdl.handle.net/10553/35376
ISSN: 1383-7621
DOI: 10.1016/j.sysarc.2017.08.002
Source: Journal of Systems Architecture[ISSN 1383-7621],v. 80, p. 30-40
Appears in Collections:Artículos
Show full item record

SCOPUSTM   
Citations

16
checked on May 9, 2021

WEB OF SCIENCETM
Citations

12
checked on May 9, 2021

Page view(s)

9
checked on May 10, 2021

Google ScholarTM

Check

Altmetric


Share



Export metadata



Items in accedaCRIS are protected by copyright, with all rights reserved, unless otherwise indicated.