Estudio comparativo de implementaciones paralelas para el análisis de imágenes obtenidas de forma remota
Loading...
Download
Official URL
Full text at PDC
Publication date
2026
Defense date
24/10/2025
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Complutense de Madrid
Citation
Abstract
Este trabajo de Tesis Doctoral aborda el desafío del desmezclado espectral en imágenes hiperespectrales mediante una solución optimizada para plataformas FPGA, utilizando el lenguaje de alto nivel DPC++ dentro del entorno Intel oneAPI. La principal dificultad en el análisis de imágenes hiperespectrales radica en la mezcla espectral de materiales dentro de un único píxel, lo que dificulta la identificación delos componentes puros o endmembers. Esta tarea es esencial para diversas aplicaciones de observación remota de la Tierra, tales como la agricultura, la minería y la gestión ambiental.La investigación presenta una cadena de procesamiento para desmezclar espectralmente los píxeles mistos de una imagen hiperespectral, que consta de tres etapas principales: 1) estimación del número de endmembers, 2) detección y clasificación automática de los materiales puros, y 3) estimación de la proporción de cada endmember en la mezcla. El proceso ha sido optimizado para su implementación en FPGA, lo que permite una reducción significativa en los tiempos de procesamiento...
This doctoral thesis addresses the challenge of spectral unmixing in hyperspectral images throug an optimized solution for FPGA platforms, using the high-level programming language DPC++ within the Intel oneAPI environment. The mail difficulty in hyperspectral image analysis lies in the spectral mixing of materials within a single pixel, which hinders the identification of pure components or endmembers. This task is essential for various Earth observation applications, such as agruculture, mining, and environmental management.The research presents a processing chain for spectral unmixing of mixed pixels in a hyperspectral image, which consists of three main stages: 1) estimation of the number of endmembers, 2) automatic detection and classification of pure materials, and 3) estimation of the proportion of each endmember in the mixture. The process has been optimized for FPGA implementation, which allows for a significant reduction in processing times...
This doctoral thesis addresses the challenge of spectral unmixing in hyperspectral images throug an optimized solution for FPGA platforms, using the high-level programming language DPC++ within the Intel oneAPI environment. The mail difficulty in hyperspectral image analysis lies in the spectral mixing of materials within a single pixel, which hinders the identification of pure components or endmembers. This task is essential for various Earth observation applications, such as agruculture, mining, and environmental management.The research presents a processing chain for spectral unmixing of mixed pixels in a hyperspectral image, which consists of three main stages: 1) estimation of the number of endmembers, 2) automatic detection and classification of pure materials, and 3) estimation of the proportion of each endmember in the mixture. The process has been optimized for FPGA implementation, which allows for a significant reduction in processing times...
Description
Tesis inédita de la Universidad Complutense de Madrid, Facultad de Informática, Departamento de Arquitectura de Computadores y Automática, leída el 24 de octubre de 2025













