Three related electromagnetic scattering problems, namely, high frequency electromagnetic (EM) ray tracing, scattering feature extraction, and inverse scattering are studied in this dissertation. New approaches are presented to advance the state of the art in each of the areas. The presented study in electromagnetic ray tracing leads to an alternative ray tracing algorithm which can outperform the traditional algorithms for complex targets. The performance of the proposed techniques demonstrates their potential application to the study of high-frequency EM scattering prediction. Second, a genetic algorithm (GA)-based algorithm with an adaptive-feeding technique is developed to simultaneously extract both scattering centers and resonances. Scattering feature extraction algorithms are then developed with the consideration of the visibility of scattering centers. Inverse scattering problems with strong multiple scattering effects are also studied. A GA-based method is presented to invert the shapes with multiple scattering effects. An approach combining hybrid GA with the tabu list idea are then developed to further improve the performance of the GA-based inversion algorithms. / text

LIDAR es un acrónimo de LIght Detection And Ranging. En la presente tesis, se usan técnicas basadas en lidar elástico para monitorizar la atmósfera remotamente y derivar información cuantitativa acerca de sus parámetros ópticos. Esta tesis doctoral comprende el diseño y operación de una estación lidar elástica basada en un láser pulsado de Nd:YAG operando a las longitudes de onda de 1064 y 532 nm, en lo que se refiere a los sistemas de recepción, control y diseño de algoritmos de inversión. Básicamente, puede dividirse en tres partes bien diferenciadas: La primera (Caps. 1, 2 y 3) comprende el estudio de la dispersión elástica (Rayleigh y Mie) en la atmósfera, orientada al cálculo del balance de enlace, e intenta vislumbrar la interrelación entre variables físicas tales como la temperatura, la presión y la humedad, y el fenómeno de dispersión, dejando de lado su posible extrapolación a modelos meteorologicos. Partiendo de esta base, se estiman valores de extinción y retrodispersión para diferentes condiciones atmosféricas y, como resultado, se presenta un balance de enlace para el sistema. El mismo incluye el estudio del alcance lidar, la estimación de la relación señal a ruido, y la evaluación de fotodiodos para diferentes librerías del usuario. Esta primera parte se cierra con las especificaciones globales del sistema. La segunda parte de este trabajo (Caps. 4, 5 y 6) atiende al diseño e implemen-tación del receptor, sistemas de sincronización y control. El receptor optoelectrónico se basa en amplificadores realimentados en corriente y cuenta con un excelente producto ganancia ancho de banda. Por lo que respecta al subsistema de sincronismo, se presentan dos unidades distintas con vistas a un futuro sistema lidar de escaneo, lo cuál ofrece la posibilidad de realizar scans entrelazados. Para terminar, el sistema de control diseñado se basa en el software de control LabView, que ofrece una filosofía de control distribuido. Con este propósito, se han especificado e implementado protocolos de bus lidar y su señalización para la presente estación lidar. Finalmente, la tercera parte comprende el diseño de algoritmos de inversión con y sin memoria (Caps. 7 y 8). Los algortimos sin memoria para atmósferas homogéneas se basan en procedimientos de ajuste por regresión como son el método de la pendiente y el de mínimos cuadrados y, en el caso de atmósferas inhomogéneas, se basan en el método de Klett y calibraciones adecuadas. Los algortimos con memoria se basan en diferentes modelos estocásticos para la atmósfera y filtrado de Kalman no lineal. Además de los algortimos de inversión, también se calculan y discuten las curvas del error de inversión. El Cap. 9 describe las medidas llevadas a cabo con el sistema que este trabajo ha permitido construir así como el resultado de aplicar los algoritmos de inversión presentados en los capítulos precedentes. La inversión de escenas reales comprende estudios de la estructura de polución, estudios de nubes (ceilometría, básicamente desplazamiento y estructura de las nubes) y señala posibles fuentes de error en el factor de solapamiento. / LIDAR is an acronym of LIght Detection And Ranging. In the present case, the elastic lidar techniques are used to remotely sense the atmosphere and to derive quantitative information about its optical parameters.This thesis comprises the design and operation of an elastic lidar station based on a pulsed Nd:YAG laser operating at the 1064- and 532-nm wavelengths, in the parts concerning receiver, control systems, and inversion algorithms.Basically, it can be divided in three different parts: The first one (Chaps. 1, 2, and 3) encompasses the study of the elastic scattering (Rayleigh and Mie) in the atmosphere for link-budget purposes and gives some insight into the interweaving between physical variables such as temperature, pressure and humidity, and the scattering phenomena, letting apart any possible extrapolation to meteorological models. From this basis, extinction and backscatter figures for different atmospheric conditions can readily be assessed and, as result, a system link budget is presented. This includes lidar range study, signal-to-noise ratio assessment, and photodiode evaluation from custom-made libraries. At the end of the first part, the system specification is made. The second part of this work (Chaps. 4, 5, and 6) is concerned with the design and implemen-tation of receiver, synchronization, and control systems. The optoelectronic receiver is based on current-feedback amplifiers and features a very large gain-bandwidth product. As for the synchronization subsystem, two different units are presented with a view to a future scanning lidar system, which makes room for interspersed scans. Eventually, the control system designed is LabView based and features a distributed control philosophy. For that purpose, lidar bus protocols and signals are specified and built for the actual lidar station. Finally, the third part encircles the design of inversion algorithms with and without memory (Chaps. 7 and 8). Non-memory algorithms for homogeneous atmospheres are based on regression curve-fitting procedures, such as the slope-method and the least squares while in instances of inhomogeneous atmospheres they are based on Klett's method and appropriate calibrations. Memory algorithms are based on different stochastic models for the atmosphere and on non-linear Kalman filtering. In addition to these inversion procedures, error assessment plots are also derived and discussed. Chap. 9 describes the measurements carried out with the system this work has contributed to build and the results of applying to them the inversion algorithms discussed in the preceding chapters.The inversion of live-scenes involves pollution structure studies, cloud studies (ceilometry, cloud motion and wave clouds, basically), and hints overlap factor error sources.

The objective of this thesis work is to detect and to characterize three-dimensional targets in a disordered medium, using electromagnetic excitations. This research domain is of great interest in many applications, such as subsoil probing, medical imaging, non-destructive testing and geophysical exploration, etc. In order to extract the target information from the heterogeneities of the medium, we propose to use one of the time reversal technique, namely the DORT method (French acronym for Décomposition de l'Opérateur de Retournement Temporel). This method permits us to generate different waves that focus selectively on each target in high noisy environment. Moreover, this method is also combined with a non-linear inversion algorithm, which permits not only to localize but also to characterize the targets. The reconstruction resolution appears to be better than the ones obtained with the DORT or the inversion procedure alone, especially in the illumination direction. It is also shown that using full-polarized data is indispensable for achieving better performances rather than in scalar configuration. Moreover, in the half-space configuration, it is mandatory to use the frequency-diversity data to get an accurate reconstruction. These theoretical developments are also confronted to experimental data measured in the optical domain. A full-polarization Tomographic Diffractive Microscopy (TDM) is implemented and a resolution about one-fourth of the wavelength is thus obtained. Furthermore, the DORT method is applied in TDM to realize selective focalization and characterization. In the presence of multiple targets, selective characterization of each scatterer is achieved.This thesis work also deals with the characterization problem using transient data. Different inversion algorithms are validated using synthetic and experimental hyper-frequency data.

L'objectif de ce travail de thèse est de détecter et de caractériser des cibles tridimensionnelles dans un milieu désordonné. Ce domaine de recherche est d'intérêt pour de nombreuses applications, telles que le sondage du sous-sol, l'imagerie médicale, la détection non-destructive et l'exploration géophysique, etc. Afin de distinguer les cibles des hétérogénéités du milieu, nous proposons d'utiliser l'une des techniques de retournement temporel, à savoir la méthode DORT (Décomposition de l'Opérateur de Retournement Temporel). La méthode DORT permet de générer des ondes focalisant sélectivement sur chaque cible présente dans un environnement fortement hétérogène. Par ailleurs, la richesse de ces ondes focalisantes est combinée avec un algorithme d'inversion non-linéaire. Ceci nous permet non seulement de localiser, mais aussi de caractériser les cibles (forme et permittivité). La résolution obtenue à l'aide de cette approche est bien meilleure que celles obtenues avec la méthode DORT ou la méthode d'inversion seules, en particulier dans la direction d'illumination. Cette résolution est d'autant meilleure que les données utilisées sont vectorielles. Dans le cas spécifique d'une configuration d'objets enfouis impliquant deux semi-espaces infinis, la caractérisation s'avère problématique. Une solution est apportée en appliquant l'approche de marche récurrente en fréquences. Ces développements théoriques sont également confrontés aux données expérimentales mesurées dans le domaine optique. Une nouvelle Microscopie Tomographique par Diffraction (MTD) est mise en œuvre dans le cadre de cette thèse en tenant compte du caractère vectoriel de la lumière. Ce faisant, une résolution d'environ un quart de la longueur d'onde a été obtenue sur des échantillons en résine déposés sur un substrat de silicium. De plus, nous avons aussi appliqué avec succès la méthode DORT à la MTD afin de focaliser et caractériser de manière sélective plusieurs diffuseurs de tailles différentes.Lors de ce travail de thèse nous avons également développé des méthodes de caractérisation en régime transitoire. Les différentes méthodes d'inversion élaborées dans ce cadre ont été validées sur des données synthétiques et expérimentales dans le domaine des radio-fréquences. / The objective of this thesis work is to detect and to characterize three-dimensional targets in a disordered medium, using electromagnetic excitations. This research domain is of great interest in many applications, such as subsoil probing, medical imaging, non-destructive testing and geophysical exploration, etc. In order to extract the target information from the heterogeneities of the medium, we propose to use one of the time reversal technique, namely the DORT method (French acronym for Décomposition de l'Opérateur de Retournement Temporel). This method permits us to generate different waves that focus selectively on each target in high noisy environment. Moreover, this method is also combined with a non-linear inversion algorithm, which permits not only to localize but also to characterize the targets. The reconstruction resolution appears to be better than the ones obtained with the DORT or the inversion procedure alone, especially in the illumination direction. It is also shown that using full-polarized data is indispensable for achieving better performances rather than in scalar configuration. Moreover, in the half-space configuration, it is mandatory to use the frequency-diversity data to get an accurate reconstruction. These theoretical developments are also confronted to experimental data measured in the optical domain. A full-polarization Tomographic Diffractive Microscopy (TDM) is implemented and a resolution about one-fourth of the wavelength is thus obtained. Furthermore, the DORT method is applied in TDM to realize selective focalization and characterization. In the presence of multiple targets, selective characterization of each scatterer is achieved.This thesis work also deals with the characterization problem using transient data. Different inversion algorithms are validated using synthetic and experimental hyper-frequency data.

O desenvolvimento de algoritmos computacionais para a obtenção de distribuições de tamanho de partícula em dispersões e que utilizam dados espectroscópicos em tempo real e in-line a partir de sensores permitirá uma variedade de aplicações, como o monitoramento de propriedades em fluidos de corte industriais, acompanhamento de processos de polimerização, tratamento de efluentes e sensoriamento atmosférico. O presente estudo tem como objetivo a implementação e comparação de técnicas para resolução de problemas de inversão, desenvolvendo algoritmos que forneçam distribuição de tamanho de partículas em dispersões a partir de dados de espectroscopia UV-Vis-Nir (Ultravioleta, Visível e Infravermelho próximo). Foram implementadas quatro técnicas, sendo uma delas um método alternativo sem a presença de etapas de inversão. Os métodos que utilizaram alguma técnica de inversão evidenciaram a dificuldade em se obter distribuições de tamanho de gotas (DTG) de boa qualidade, enquanto o método alternativo foi aquele que se mostrou mais eficiente e confiável. Este estudo é parte de um programa cooperativo entre a Universidade de São Paulo e a Universidade de Bremen chamado programa BRAGECRIM (Brazilian German Cooperative Research Initiative in Manufacturing) e é financiado pela FAPESP, CAPES, FINEP e CNPq (Brasil) e DFG (Alemanha). / The development of computer algorithms to obtain the particle size distributions in dispersions using spectroscopic data in real time and in-line from sensors enable a variety of applications such as monitoring properties in industrial cutting fluids, monitoring of polymerization processes, wastewater, atmospheric sensing and other applications. The aim of this study is to implement techniques for inversion problem solving, by testing algorithms that provide particle size distribution in dispersions from UV-Vis-Nir (Ultraviolet, Visible and Near-Infrared) spectroscopic data. Four techniques have been implemented, one of them being an alternative method without the inversion step. The methods using inversion techniques showed difficulties to obtain droplet size distributions (DSD) with good quality, while the alternative method was the one that was more efficient and reliable. This study is part of a cooperative program between the University of São Paulo and the University of Bremen, within the BRAGECRIM program (Brazilian German Cooperative Research Initiative in Manufacturing) and is financially supported by FAPESP, CAPES, FINEP and CNPq (Brazil) and DFG (Germany).

O desenvolvimento de algoritmos computacionais para a obtenção de distribuições de tamanho de partícula em dispersões e que utilizam dados espectroscópicos em tempo real e in-line a partir de sensores permitirá uma variedade de aplicações, como o monitoramento de propriedades em fluidos de corte industriais, acompanhamento de processos de polimerização, tratamento de efluentes e sensoriamento atmosférico. O presente estudo tem como objetivo a implementação e comparação de técnicas para resolução de problemas de inversão, desenvolvendo algoritmos que forneçam distribuição de tamanho de partículas em dispersões a partir de dados de espectroscopia UV-Vis-Nir (Ultravioleta, Visível e Infravermelho próximo). Foram implementadas quatro técnicas, sendo uma delas um método alternativo sem a presença de etapas de inversão. Os métodos que utilizaram alguma técnica de inversão evidenciaram a dificuldade em se obter distribuições de tamanho de gotas (DTG) de boa qualidade, enquanto o método alternativo foi aquele que se mostrou mais eficiente e confiável. Este estudo é parte de um programa cooperativo entre a Universidade de São Paulo e a Universidade de Bremen chamado programa BRAGECRIM (Brazilian German Cooperative Research Initiative in Manufacturing) e é financiado pela FAPESP, CAPES, FINEP e CNPq (Brasil) e DFG (Alemanha). / The development of computer algorithms to obtain the particle size distributions in dispersions using spectroscopic data in real time and in-line from sensors enable a variety of applications such as monitoring properties in industrial cutting fluids, monitoring of polymerization processes, wastewater, atmospheric sensing and other applications. The aim of this study is to implement techniques for inversion problem solving, by testing algorithms that provide particle size distribution in dispersions from UV-Vis-Nir (Ultraviolet, Visible and Near-Infrared) spectroscopic data. Four techniques have been implemented, one of them being an alternative method without the inversion step. The methods using inversion techniques showed difficulties to obtain droplet size distributions (DSD) with good quality, while the alternative method was the one that was more efficient and reliable. This study is part of a cooperative program between the University of São Paulo and the University of Bremen, within the BRAGECRIM program (Brazilian German Cooperative Research Initiative in Manufacturing) and is financially supported by FAPESP, CAPES, FINEP and CNPq (Brazil) and DFG (Germany).

L'importance des enjeux liés à la qualité et à la connaissance de l'eau souterraine nécessite de caractériser le fonctionnement des nappes d’eaux souterraines. En hydrogéologie, les méthodes qui permettent de réaliser le modèle conceptuel d’un aquifère reposent généralement sur l’observation et la caractérisation du milieu, à partir notamment de la mise en place de forages et la réalisation de pompages d’essai. Cependant, ces méthodes présentent des limites pour caractériser la variabilité spatiale des aquifères hétérogènes. Dans cette thèse, nous nous sommes intéressés au développement d’approches expérimentales qui utilisent des informations issues de traçage thermique ou salin pour la caractérisation des paramètres hydrauliques des milieux hétérogènes poreux. Ce choix expérimental a été justifié 1) par les difficultés de prise en compte des hétérogénéités hydrauliques dans les milieux poreux naturels et 2) pour une meilleure compréhension des phénomènes mis en jeu en situation contrôlée. Dans les travaux réalisés, nous avons tout d’abord proposé d’utiliser des mesures thermiques ponctuelles, puis obtenues dans l'infrarouge thermique pour estimer les paramètres hydrauliques en milieux poreux hétérogènes. Cependant, les méthodes de traçage thermiques étant limitées dans les zones faiblement perméables, nous avons fait évoluer notre approche en la combinant avec une méthode géophysique (Potentiel Spontané). Cette méthode, sensible aux écoulements des fluides et aux processus thermoélectrique et électrochimique, nous a permis de collecter des données temporelles intéressantes pour surveiller des variations thermiques ou saline provoquées dans les milieux. Les mesures obtenues ont été par la suite utilisées dans des algorithmes d’inversion pour estimer les distributions spatiales des propriétés hydrauliques. Les résultats obtenus nous ont permis de montrer l’efficacité de ces approches pour la caractérisation des milieux hétérogènes et par conséquent pour la modélisation des écoulements des fluides dans ces milieux. / The importance of issues related to the quality of groundwater requires characterizing the operation of groundwater aquifers. Hydrogeological methods used to carry out conceptual model of an aquifer are generally based on the observation and characterization of the medium, by using boreholes and carrying out of pumping test operations. However, these methods present some limitations in characterizing the spatial variability of heterogeneous aquifers. In this thesis, we develop experimental approaches that use information from thermal or saline tracing for the characterization of hydraulic parameters in heterogeneous porous media. The choice of using experimental approaches was justified 1) by difficulties of taking into account hydraulic heterogeneities in natural porous media and 2) for a better understanding of the phenomena involved in a controlled situation. In this work, we first proposed the use of punctual thermal measurements, then measurements obtained by using thermal infrared to estimate hydraulic parameters in heterogeneous porous media. However, as thermal tracing methods are limited in low permeability areas, we improved our approach by combining it with a geophysical method (Spontaneous Potential). This method, sensitive to fluid flows and thermoelectric and electrochemical processes, has allowed us to collect interesting temporal data to monitor thermal or saline variations in the media. The measurements obtained were then used in inversion algorithms to estimate the spatial distributions of hydraulic properties. The results obtained allowed to demonstrate the effectiveness of these approaches for the characterization of heterogeneous media and consequently for the modelling of fluid flows.