Show simple item record

dc.contributor.advisorRivera-Gallego, Wilson
dc.contributor.authorLugo-Beauchamp, Wilfredo E.
dc.description.abstractHyperspectral imaging provides the capability to identify and classify materials remotely. The applications of such technology is applied everywhere from medical devices and military targets to environmental sciences. With the ongoing advances in spectrometers (spatial resolution and bits per pixel density) the data gathered is constantly increasing. Some hyperspectral imaging algorithms could easily take days or weeks in analyzing a full single hyperspectral data set. In this thesis we performed a porting and parallelization of four hyperspectral algorithms representative of the type of analysis done in a typical data set. Two of the algorithms are in the area of data classication, one in the area of feature reduction and the other one is a combination of both areas. The parallelized algorithms were benchmarked on the Intel 32 bits Pentium M architecture and the new Intel 64 bits Itanium 2 architecture. For three of the four algorithms we demonstrated that the use of parallel approaches in combination with computational clusters speedup signicantly the executions times and provide great scalability. On the other algorithm, based on linear algebra manipulations using distributed objects, we obtained execution times that took longer than the sequential implementation. A systematic performance analysis is carried out to explain the performance behavior of the algorithms.en_US
dc.description.abstractLa capacidad de analizar imagenes hiper-espectrales provee la habilidad de indentificar y clasificar materiales remotamente. Las aplicaciones de este tipo de tecnología tiene aplicaciones en un gran sinnúmero de áreas que van desde aparatos médicos y objectivos militares a ciencias ambientales. Debido a los continuos avances en los sensores espectrales (resolución espacial y en la cantidad de bits en un pixel) la cantidad de data recojida esta aumentando constantemente. Algoritmos hiper-espectrales pueden tomar das e incluso semanas en analizar todas las bandas de una muestra. Como parte de esta tesis portamos y paralelizamos 4 algoritmos hiper-espectrales representativos del tipo de analsis efectuado en una imagen hiper-espectral comunmente. Dos de los algoritmos son basados en classificadores, uno en el área de redución de bandas y el restante es una combinación de ambas áreas. Los algoritmos paralelizados fueron probados en las arquitecturas de Intel Pentium M (32 bits) e Intel Itanium 2 (64 bits). En tres de los cuatro algoritmos quedo demostrado que la paralelización de los algoritmos proveen tiempos de ejecución mucho mas rápidos y con una gran escalabilidad. En el algoritmo restante, basado en manipulaciones de algebra lineal y objectos distribuidos, los tiempos de ejecución resultaron ser mayores que los de la implementación secuencial. Un análisis sistemático de eficiencia es llevado a cabo para explicar el comportamiento de crecimiento computacional de los algoritmos.en_US
dc.subjectHyperspectral imaging classificationen_US
dc.subjectdimensionality reduction algorithmsen_US
dc.titleParallelization of hyperspectral imaging classification and dimensionality reduction algorithmsen_US
dc.rights.licenseAll rights reserveden_US
dc.rights.holder(c) 2004 Wilfredo E. Lugo-Beauchampen_US
dc.contributor.committeeHunt, Shawn
dc.contributor.committeeSeguel, Jaime
dc.contributor.representativeVasquez, Pedro Engineeringen_US
dc.contributor.collegeCollege of Engineeringen_US
dc.contributor.departmentDepartment of Electrical and Computer Engineeringen_US

Files in this item


This item appears in the following Collection(s)

  • Theses & Dissertations
    Items included under this collection are theses, dissertations, and project reports submitted as a requirement for completing a graduate degree at UPR-Mayagüez.

Show simple item record

All rights reserved
Except where otherwise noted, this item's license is described as All Rights Reserved