A supervised classification system of spectral images acquired with a single pixel optical architecture and side information

Simple item page
dc.contributor.advisorArguello Fuentes, Henry
dc.contributor.authorSanchez Quiroga, Karen Yaneth
dc.date.accessioned2023-04-06T20:23:07Z
dc.date.available2023
dc.date.available2023-04-06T20:23:07Z
dc.date.created2019
dc.date.issued2019
dc.description.abstractLas imagenes espectrales proporcionan una gran cantidad de informaci ´ on que permite re- ´ alizar diversas tareas de procesamiento, como clasificacion, con gran precisi ´ on. Sin embargo, ´ debido a la alta dimensionalidad de los datos, procesar, transmitir y almacenar dicha informacion es costoso. En los ´ ultimos a ´ nos, la compresi ˜ on de im ´ agenes espectrales (CSI) ha ´ emergido como una nueva tecnica de adquisici ´ on que adquiere proyecciones codificadas de ´ la escena espectral aplicando diferentes patrones de codificacion, reduciendo considerable- ´ mente los costos de almacenamiento y transmision. Variando la estrategia de muestreo, varios ´ dispositivos CSI, con diferentes configuraciones opticas, se han desarrollado, donde la arqui- ´ tectura de camara de un solo p ´ ´ıxel (SPC) sobresale por bajo costo de implementacion. Tradi- ´ cionalmente, una reconstruccion completa de la escena subyacente es necesaria para realizar ´ cualquier tarea de procesamiento, lo que implica resolver un problema de optimizacion com- ´ putacionalmente costoso. El objetivo de este proyecto es realizar clasificacion de im ´ agenes ´ espectrales utilizando directamente mediciones CSI, evitando la reconstruccion completa de ´ la escena. Las mediciones de CSI se adquiriran mediante la implementaci ´ on en el laboratorio ´ de una SPC. Ademas, dada la baja resoluci ´ on espacial del sensor SPC, se propone obtener ´ informacion complementaria a trav ´ es de un sensor RGB auxiliar, que tiene una resoluci ´ on es- ´ pacial mas alta. Utilizando la informaci ´ on de ambos sensores, este trabajo propone dise ´ nar los ˜ patrones de codificacion SPC considerando la agrupaci ´ on de p ´ ´ıxeles con caracter´ısticas similares en la imagen RGB. Luego, es posible extraer caracter´ısticas de la escena para realizar una clasificacion directa. Por lo tanto, es posible obtener un mapa de clasificaci ´ on, utilizando ´ una maquina de soporte vectorial de manera r ´ apida y con alta precisi ´ on sin requerir una etapa ´ de reconstruccion. En general, se obtuvo una precisi ´ on global de ´ 95.41%, 97.29%, 97.72% y 99% utilizando la “Pavia University”, “Pavia Center”, “Salinas”, y “granos de cacao” adquiridos en un laboratorio optico, respectivamente.
dc.description.abstractenglishSpectral images provide a large amount of information allowing to perform different processing tasks, such as classification, with greater precision. However, due to the data highdimensionality, processing, transmitting, and storing such information is expensive. In recent years, Compressive Spectral Imaging (CSI) has emerged as a new acquisition technique which acquires coded projections of the spectral scene by applying different coding patterns, considerably reducing storage and transmission costs. Varying the sampling strategy, several CSI devices, with different optical configurations, have been developed, where the single pixel camera architecture (SPC) excels with a low implementation cost. Traditionally, a complete reconstruction of the underlying scene is needed to perform any processing task, which involves solving a computationally expensive optimization problem. The objective of this project is to perform spectral image classification by directly using the CSI measurements, thus avoiding the complete reconstruction of the scene. The CSI measurements will be acquired through the implementation in the laboratory of a SPC. Besides, given the low spatial resolution of the SPC sensor, it is proposed to obtain side information through an auxiliary RGB sensor, which has a higher spatial resolution. Using the information from both sensors, this work proposes to design the SPC coding patterns considering the grouping of pixels with similar characteristics in the RGB image. Then, it is possible to extract valuable features of the scene to perform a direct classification. Hence, it is possible to obtain a classification map, using a support vector machine, in a fast way and with high accuracy in the classification without requiring a reconstruction stage. In general, a global precision of 95.41%, 97.29%, 97.72% and 99% was obtained using the “Pavia University”, “Pavia Center”, “Salinas”, and spectral images of “cocoa beans” acquired in an optical laboratory, respectively
dc.description.degreelevelMaestría
dc.description.degreenameMagíster en Ingeniería Electrónica
dc.format.mimetypeapplication/pdf
dc.identifier.instnameUniversidad Industrial de Santander
dc.identifier.reponameUniversidad Industrial de Santander
dc.identifier.repourlhttps://noesis.uis.edu.co
dc.identifier.urihttps://noesis.uis.edu.co/handle/20.500.14071/14020
dc.language.isospa
dc.publisherUniversidad Industrial de Santander
dc.publisher.facultyFacultad de Ingenierías Fisicomecánicas
dc.publisher.programMaestría en Ingeniería Electrónica
dc.publisher.schoolEscuela de Ingenierías Eléctrica, Electrónica y Telecomunicaciones
dc.rightsinfo:eu-repo/semantics/openAccess
dc.rights.accessrightsinfo:eu-repo/semantics/openAccess
dc.rights.coarhttp://purl.org/coar/access_right/c_abf2
dc.rights.creativecommonsAtribución-NoComercial-SinDerivadas 4.0 Internacional (CC BY-NC-ND 4.0)
dc.rights.licenseAttribution-NonCommercial 4.0 International (CC BY-NC 4.0)
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/
dc.subjectImagenes Espectrales
dc.subjectResoluci ´ On Espacial
dc.subjectMuestreo Compresivo
dc.subjectSen- ´ Sado Espectral De Imagenes
dc.subjectFusi ´ On De Im ´ Agenes
dc.subjectClasificaci ´ On Supervisada
dc.subjectSuperpixeles.
dc.subject.keywordSpectral Imaging
dc.subject.keywordSpatial Resolution
dc.subject.keywordCompressive Sensing
dc.subject.keywordCompressive Spectral Imaging
dc.subject.keywordImage Fusion
dc.subject.keywordSupervised Classification
dc.subject.keywordSuperpixels.
dc.titleA supervised classification system of spectral images acquired with a single pixel optical architecture and side information
dc.title.englishA supervised classification system of spectral images acquired with a single pixel optical architecture and side information. *
dc.type.coarhttp://purl.org/coar/version/c_b1a7d7d4d402bcce
dc.type.hasversionhttp://purl.org/coar/resource_type/c_bdcc
dc.type.localTesis/Trabajo de grado - Monografía - Maestría
dspace.entity.type
Files
Original bundle
Now showing 1 - 3 of 3
No Thumbnail Available
Name:
Carta de autorización.pdf
Size:
266.66 KB
Format:
Adobe Portable Document Format
Download
No Thumbnail Available
Name:
Documento.pdf
Size:
56.61 MB
Format:
Adobe Portable Document Format
Download
No Thumbnail Available
Name:
Nota de proyecto.pdf
Size:
201.35 KB
Format:
Adobe Portable Document Format
Download
Collections
Maestría en Ingeniería Electrónica