Publication: Contribución a la solución del problema inverso en la detección remota de gases y partículas mediante radiometría espectral y de imagen en el infrarrojo
Loading...
Identifiers
Publication date
2016-01
Defense date
2016-02-11
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Esta Tesis tiene como objetivo último demostrar las ventajas de los sistemas multiespectrales en problemas complejos, en la resolución del problema inverso en la detección remota, mediante sensores IR, de compuestos o propiedades de interés, que de otra manera no serían unívocos. Para ello, se ha empleado una metodología basada en el modelado y la simulación del problema radiométrico como herramienta para la comprensión del problema y la posterior aplicación de técnicas de medida, resolución del problema inverso y recuperación de datos. El carácter espectral de la señal utilizada proporciona, una gran robustez al sistema, con los dos parámetros característicos que definen la calidad de un sistema global de detección, la Probabilidad de Detección (PD) y la Tasa de Falsa Alarma (TFA). La selección óptima de las bandas es un tema clave en este problema. Los resultados de investigaciones previas a este trabajo en el LIR-Infrared Lab, indican que en general hay un número bajo de Longitudes de Onda que resuelven óptimamente un problema concreto y que usar un número mayor puede generar problemas de sobredimensionamiento que reducen las prestaciones globales del sistema en la PD y la PFA.
Las importantes capacidades de los sistemas espectralmente optimizados, se han demostrado en este trabajo a través de dos aplicaciones que dan soluciones específicas a problemas muy diferentes entre sí. Una de ellas, es la relativa a sensores de imagen infrarroja multiespectral para detectar las emisiones de los vehículos más contaminantes a través del estudio radiativo de la pluma de gases de escape. La otra es un sensor infrarrojo bi-espectral y bi-direccional de baja masa para el estudio in-situ del polvo en suspensión en la superficie de Marte.
El denominador común de estas aplicaciones radica en que ambas presentan unas características espectrales muy definidas. Ambos problemas han podido ser modelados radiométricamente lo que hace posible simular los escenarios potenciales y definir la mejor solución dentro de los límites fijados. Finalmente, las dificultades en ambos casos también son del mismo tipo: su resolución por métodos convencionales plantea un problema de indeterminación y es que existen infinitas combinaciones de variables que intervienen en el proceso de transferencia radiativa y que arrojan el mismo resultado directo, es decir una secuencia temporal, organizada espacialmente, de mV proporcionados por los detectores.
Por tanto, es necesario y ese ha sido nuestra aportación fundamental en este trabajo de tesis, proponer metodologías que permitan resolver con la suficiente precisión y dentro de los límites fijados, el problema inverso de manera unívoca. Es decir, a partir de los datos (mV) proporcionados por los detectores, como respuesta a la evolución temporal de las propiedades y características de la nube de gas emitida por un vehículo o de la distribución de partículas en una cierta región de Marte, obtener la concentración de gases o partículas, e incluso su temperatura, de manera unívoca.
The objective of this thesis is to demonstrate the advantages of multispectral systems in the resolution of the inverse problem in remote detection, by means of infrared sensors, of compounds or properties of interest, which otherwise are not univocal. A methodology has been used based on the modelling and simulation of the radiometric problem as a tool to understand the problem later apply techniques of measurement, inverse problem resolution and data recovery. The spectral nature of the used signal provides a great robustness to the system regarding both typical parameters that define the quality of a global detection system, the Probability of Detection (PD) and the false alarm rate (FAR). The optimal selection of spectral bands is a key theme in this problem. Previous research results in the LIR-Infrared Lab indicate that in general a small number of wavelengths can be found that solve optimally a specific problem and that the use of a higher number of bands can lead to oversize problems that reduce global system capabilities (reflected in a deterioration of the PD and FAR). The capabilities of the spectrally optimized systems have been demonstrated in this work by means of two applications that give specific solutions to very different problems. One of them is related to the use of infrared multispectral imaging to detect and measure the gas emission of highly pollutant vehicles by the spectral study of the escape gases plume. The other one is an infrared bi-spectral and bidirectional sensor of low mass for the in-situ study of powder in suspension on the Martian atmosphere. The common denominator of these applications lies in that both of them show very defined spectral characteristics. Both problems have been modelled radiometrically and therefore it is possible to simulate the potential scenarios and to define the best solution inside the fixed limits. Finally, the difficulties in both cases also are of the same type: their resolution by conventional methods raises a problem of indetermination since there exists an infinite number of combinations of variables that intervene in the process of radiative transfer giving rise to the same direct result, i.e. a temporal sequence, spatially organized, of mV provided by the detectors. For this reason, it is necessary, and it has been our main fundamental contribution in this theses work, to propose methodologies that allow resolving with sufficient precision and inside fixed limits, the inverse problem in a univocal way. That is, starting with the data (mV) provided by the detectors, as a response to the temporal evolution of the properties of a gas cloud emitted by either a pollutant vehicle or a dust cloud on the Martian surface, to obtain the gas or particle concentration, even temperature, in a univocal way.
The objective of this thesis is to demonstrate the advantages of multispectral systems in the resolution of the inverse problem in remote detection, by means of infrared sensors, of compounds or properties of interest, which otherwise are not univocal. A methodology has been used based on the modelling and simulation of the radiometric problem as a tool to understand the problem later apply techniques of measurement, inverse problem resolution and data recovery. The spectral nature of the used signal provides a great robustness to the system regarding both typical parameters that define the quality of a global detection system, the Probability of Detection (PD) and the false alarm rate (FAR). The optimal selection of spectral bands is a key theme in this problem. Previous research results in the LIR-Infrared Lab indicate that in general a small number of wavelengths can be found that solve optimally a specific problem and that the use of a higher number of bands can lead to oversize problems that reduce global system capabilities (reflected in a deterioration of the PD and FAR). The capabilities of the spectrally optimized systems have been demonstrated in this work by means of two applications that give specific solutions to very different problems. One of them is related to the use of infrared multispectral imaging to detect and measure the gas emission of highly pollutant vehicles by the spectral study of the escape gases plume. The other one is an infrared bi-spectral and bidirectional sensor of low mass for the in-situ study of powder in suspension on the Martian atmosphere. The common denominator of these applications lies in that both of them show very defined spectral characteristics. Both problems have been modelled radiometrically and therefore it is possible to simulate the potential scenarios and to define the best solution inside the fixed limits. Finally, the difficulties in both cases also are of the same type: their resolution by conventional methods raises a problem of indetermination since there exists an infinite number of combinations of variables that intervene in the process of radiative transfer giving rise to the same direct result, i.e. a temporal sequence, spatially organized, of mV provided by the detectors. For this reason, it is necessary, and it has been our main fundamental contribution in this theses work, to propose methodologies that allow resolving with sufficient precision and inside fixed limits, the inverse problem in a univocal way. That is, starting with the data (mV) provided by the detectors, as a response to the temporal evolution of the properties of a gas cloud emitted by either a pollutant vehicle or a dust cloud on the Martian surface, to obtain the gas or particle concentration, even temperature, in a univocal way.
Description
Keywords
Sensores de imagen infrarroja multiespectral, Sensores IR, Detección remota, Problema inverso