[pt] O tratamento de ambientes contaminados e a elaboração de programas de monitoramento requerem estimativas dos níveis de concentrações das substâncias,bem como um conhecimento de seu destino, taxa de transporte e tempos de residência. Na ausência de dados obtidos através de monitoramento, modelos ambientais constituem freqüentemente o único modo capaz de prever as concentrações aproximadas dos contaminantes nos compartimentos (meios) que formam o ambiente. A aplicação do modelo ambiental de multimeios Quantitative Water Air Sediment Interaction (QWASI) a um determinado ambiente exige conhecer parâmetros relacionados à área em estudo e aos contaminantes. Para o caso da Baía de Guanabara, muitos desses parâmetros não estão disponíveis. Desse modo, eles serão aproximados com técnicas do problema inverso a partir das concentrações de contaminantes medidas no sedimento e no mexilhão perna-perna. Entre os parâmetros estimados estão as emissões de 10 hidrocarbonetos policíclicos aromáticos (HPAs) nas águas da Baía de Guanabara. Este trabalho também apresenta simulações da evolução da concentração de alguns hidrocarbonetos em função do tempo considerando uma emissão inicial de um HPA em um ponto arbitrário da Baía de Guanabara. / [en] The treatment of polluted environments and the elaboration of monitoring programs request estimates of the levels of concentrations of the substances, as well as a knowledge of its destiny, rates of transport and times of residence. In the absence of data obtained through monitoring, environmental models often provide the only way to predict the approximate concentrations of the pollutants in the compartments (media) that form the environmental. The application of the multimedia environmental model Quantitative Water Air Sediment Interaction (QWASI) to a certain environment demands the knowledge of parameters related to the area in study and to the pollutants. For the case of the Guanabara Bay, many of those parameters are not available. Therefore, they will be approximated with inverse problem techniques starting from the measured concentrations of pollutants in the sediment and in the mussel perna-perna. Among the estimated parameters are the emissions of 10 polycyclic aromatic hydrocarbons (PAHs) in the waters of the Guanabara Bay. This work also presents simulations of the evolution of the concentration of some hydrocarbons infunction of the time considering an initial emission of a PAH in an arbitrary point of the Guanabara Bay.

Este trabalho trata da modelagem de sinais ultrassônicos gerados por transdutores circulares (plano e côncavo) no modo de pulso-eco, inspecionando uma região de interesse predeterminada. Para essa análise, dois modelos da resposta impulsiva de um transdutor piezelétrico foram implementados: um do transdutor circular plano e outro do transdutor circular côncavo focalizado. Este último sendo o modelo proposto pelo autor com uma geometria baseada em anéis concêntricos como elementos de área do emissor. A adição de diversos anéis concêntricos, deslocados ao longo de seu eixo e de raios sucessivamente menores, permitiu calcular o campo acústico, gerado por uma abertura côncava, bem como seu eco refletido. A resposta impulsiva de cada anel resulta da diferença entre as respostas impulsivas calculadas para um emissor circular grande e um pequeno. O modelo implementado para o cálculo dos sinais de eco foi utilizado na varredura de uma região contendo um conjunto de pontos com refletividade acústica igual a 1. A reconstrução da imagem foi feita com esses sinais aplicando-lhes a técnica de regularização de Tikhonov. A qualidade das reconstruções das imagens obtidas foi avaliada e comparada a das imagens convencionais. Dentre as métricas de avaliação das imagens reconstruídas estão a influência na variação da velocidade de propagação da onda acústica no meio, a extensão e a discretização do grid e o parâmetro de regularização alfa. Todas as reconstruções foram analisadas segundo o Erro Médio Quadrático (MSE). Finalmente, ensaios experimentais foram conduzidos para a obtenção de A-scans (imagens em modo de amplitude) os quais foram inseridas no modelo teórico para a reconstrução de imagens e analisadas pelo MSE. / This work deals with the modeling of ultrasonic signals generated by circular transducers (planar and concave pistons) in pulse-echo mode, inspecting a predetermined region of interest (ROI). For this analysis, two models of the impulse response of a transducer were implemented: one using a plane piston transducer and the other, a model proposed by the author of this work, using a concave transducer with concentric rings as elements of the emitting area. The addition of several concentric rings moved along its axis allowed us to calculate the acoustic field generated by a concave opening and the echo reflected from each point in space. The impulse response of each ring represents the difference between the impulse responses calculated for a large circular transmitter and a small one. The model implemented for calculating the echo signals is used to scan a region, within a ROI, containing a set of points with acoustic reflectivity of 1. Simulations of the regions are made with these signals by applying the Tikhonov regularization method. To evaluate the quality of image reconstruction, the images are compared with the conventional images. Among the metrics to evaluate the reconstructed images are the influence of the variation of the acoustic wave propagation in the media, the grid range and discretization and the parameter of regularization alpha. All of the image reconstructions were analyzed through the Mean Square Error (MSE) criterion. Finally, experiments were conducted in order to obtain A-scans which were then re-inserted in the theoretical model to reconstruct and analyze the images.

A tomografia por sensoriamento térmico é muito utilizada em diferentes aplicações industriais, tais como a determinação de propriedades térmicas de novos materiais, o controle da produção de calor e a temperatura no processo de manufatura. Entretanto, o emprego de técnicas tomográficas em processos industriais envolvendo transferência de calor ainda carece de métodos robustos e computacionalmente eficientes. Nesse contexto, o principal objetivo deste trabalho é contribuir para o desenvolvimento de uma técnica não intrusiva de medição do coeficiente de convecção a partir de medidas externas de temperatura e fluxo de calor baseada na solução do problema térmico inverso. Para tanto é necessário resolver um problema de condução acoplado a um problema de convecção de calor. Este acoplamento ocorre através do coeficiente de convecção no interior do domínio do problema, cuja determinação pode ser feita através da aplicação de um fluxo de calor e medição das temperaturas resultantes na superfície externa. A tomografia térmica é tratada como um problema de minimização global, cuja função objetivo é um funcional de erro que quantifica a diferença entre as medidas externas não intrusivas (temperatura real) e as medidas calculadas no modelo numérico (temperatura aproximada). A natureza mal condicionada do problema assim formulado se manifesta na superfície de minimização por produzir topologias problemáticas tais como múltiplos mínimos locais, pontos de sela, vales ao redor da solução, platôs, etc. Desse modo, uma técnica bastante sofisticada, capaz de convergir para a solução correta mesmo na presença dessas patologias é necessária para obtenção da solução. Neste trabalho optou-se pelo método de Newton para a minimização deste funcional em que a inversa da matriz Hessiana é substituída por uma pseudo-inversa construída a partir da técnica de Decomposição em Valores Singulares Truncados. Os resultados mostram que a técnica proposta foi capaz de superar os problemas de convergência associados à natureza intrínseca mal condicionada do problema inverso e o coeficiente de convecção foi reconstruído com precisão razoável. / Tomography by thermal sensing is widely used in different industrial applications, such as the determination of thermal properties of new materials, the control of heat production and the temperature in manufacturing processes. However, the application of tomographic techniques in industrial processes involving heat transfer still lacks robust and computationally efficient methods. In this context, the main objective of this thesis is to contribute to the development of a non-intrusive technique for measuring of the convection coefficient from external temperature and heat flow measurements based on the solution of the inverse thermal problem. This requires solving a conduction problem coupled with a heat convection problem, which is coupled through an internal convection coefficient, determined by applying a heat flux and measuring the resulting temperatures on the external boundary. The thermal tomography is treated as a global minimization problem in which the fitness function is an error functional that quantifies the difference between non-intrusive external measurements (actual temperature) and measurements calculated in a numerical model (approximate temperature). The ill-conditioned nature of the problem manifests itself in the minimization problem for producing problematic topologies, such as multiple local minima, saddle points, valleys around the solution, plateaus, etc. Thus, a very sophisticated technique that can converge to the correct solution even in the presence of these pathologies is necessary to obtain the solution. In this thesis the Newton\'s method was used for the minimization of this functional in which the inverse Hessian matrix was replaced by a pseudo-inverse built from the truncated singular value decomposition technique. Results show that the proposed technique was capable of overcoming the convergence problems associated with the intrinsic ill-conditioned nature of the inverse problem and the convection coefficient was reconstructed within reasonable precision.

Les systèmes décrits par les équations aux dérivées partielles, appartiennent à la classe des systèmes dynamiques impliquant des fonctions dépendantes de plusieurs variables, comme le temps et l'espace. Déjà fortement répandus pour la modélisation mathématique de phénomènes physiques et environnementaux, ces systèmes ont un rôle croissant dans les domaines de l'automatique. Cette expansion, provoquée par les avancées technologiques au niveau des capteurs facilitant l'acquisition de données et par les nouveaux enjeux environnementaux, incite au développement de nouvelles thématiques de recherche. L'une de ces thématiques, est l'étude des problèmes inverses et plus particulièrement l'identification paramétrique des équations aux dérivées partielles. Tout abord, une description détaillée des différentes classes de systèmes décrits par ces équations est présentée puis les problèmes d'identification qui leur sont associés sont soulevés. L'accent est mis sur l'estimation paramétrique des équations linéaires, homogènes ou non, et sur les équations linéaires à paramètres variant. Un point commun à ces problèmes d'identification réside dans le caractère bruité et échantillonné des mesures de la sortie. Pour ce faire, deux types d'outils principaux ont été élaborés. Certaines techniques de discrétisation spatio-temporelle ont été utilisées pour faire face au caractère échantillonné des données; les méthodes de variable instrumentale, pour traiter le problème lié à la présence de bruit de mesure. Les performances de ces méthodes ont été évaluées selon des protocoles de simulation numérique reproduisant des situations réalistes de phénomènes physique et environnementaux, comme la diffusion de polluant dans une rivière / A large variety of natural, industrial, and environmental systems involves phenomena that are continuous functions not only of time, but also of other independent variables, such as space coordinates. Typical examples are transportation phenomena of mass or energy, such as heat transmission and/or exchange, humidity diffusion or concentration distributions. These systems are intrinsically distributed parameter systems whose description usually requires the introduction of partial differential equations. There is a significant number of phenomena that can be simulated and explained by partial differential equations. Unfortunately all phenomena are not likely to be represented by a single equation. Also, it is necessary to model the largest possible number of behaviors to consider several classes of partial differential equations. The most common are linear equations, but the most representative are non-linear equations. The nonlinear equations can be formulated in many different ways, the interest in nonlinear equations with linear parameters varying is studied. The aim of the thesis is to develop new estimators to identify the systems described by these partial differential equations. These estimators must be adapted with the actual data obtained in experiments. It is therefore necessary to develop estimators that provide convergent estimates when one is in the presence of missing data and are robust to measurement noise. In this thesis, identification methods are proposed for partial differential equation parameter estimation. These methods involve the introduction of estimators based on the instrumental variable technique

Prélevés de manière non invasive, les gaz respiratoires sont constitués de nombreux composés organiques volatils (VOCs) dont la quantité dépend de l’état de santé du sujet. L’analyse quantitative de l’air expiré présente alors un fort intérêt médical, que ce soit pour le diagnostic ou le suivi de traitement. Dans le cadre de ma thèse, nous proposons d’étudier un dispositif d’analyse des gaz respiratoires, et notamment de ces VOCs. Cette thèse multidisciplinaire aborde différents aspects, tels que le choix des capteurs, du matériel et des modes d’acquisition, l’acquisition des données à l’aide d’un banc gaz, et ensuite le traitement des signaux obtenus de manière à quantifier un mélange de gaz. Nous étudions la réponse d’un capteur à oxyde métallique (MOX) à des mélanges de deux gaz (acétone et éthanol) dilués dans de l’air synthétique (oxygène et azote). Ensuite, nous utilisons des méthodes de séparation de sources de manière à distinguer les deux gaz, et déterminer leur concentration. Pour donner des résultats satisfaisants, ces méthodes nécessitent d’utiliser plusieurs capteurs dont on connait la forme mathématique du modèle décrivant l’interaction du mélange avec le capteur, et qui présentent une diversité suffisante dans les mesures d’étalonnage pour estimer les coefficients de ce modèle. Dans cette thèse, nous montrons que les capteurs MOX peuvent être décrits par un modèle de mélange linéaire quadratique, et qu’un mode d’acquisition fonctionnant en double température permet de générer deux capteurs virtuels à partir d’un unique capteur physique. Pour quantifier précisément les composants du mélange à partir des mesures sur ces capteurs (virtuels), nous avons conçu des méthodes de séparation de sources, supervisées et non supervisées appliquées à ce modèle non-linéaire : l’analyse en composantes indépendantes, des méthodes de moindres carrés (algorithme de Levenberg-Marquardt), et une méthode bayésienne ont été étudiées. Les résultats expérimentaux montrent que ces méthodes permettent d’estimer les concentrations de VOCs contenus dans un mélange de gaz, de façon précise, en ne nécessitant que très peu de points de calibration. / Non-invasively taken, exhaled breath contains many volatile organic compounds (VOCs) whose amount depends on the health of the subject. Quantitative analysis of exhaled air is of great medical interest, whether for diagnosis or for a treatment follow-up. As part of my thesis, we propose to study a device to analyze exhaled breath, including these VOCs. This multidisciplinary thesis addresses various aspects, such as the choice of sensors, materials and acquisition modes, the acquisition of data using a gas bench, and then the processing of the signals obtained to quantify a gas mixture. We study the response of a metal oxide sensor (MOX) to mixtures of two gases (acetone and ethanol) diluted in synthetic air (oxygen and nitrogen). Then, we use source separation methods in order to distinguish the two gases, and to determine their concentration. To give satisfactory results, these methods require first to use several sensors for which we know the mathematical model describing the interaction of the mixture with the sensor, and which present a sufficient diversity in the calibration measurements to estimate the model coefficients. In this thesis, we show that MOX sensors can be described by a linear-quadratic mixing model, and that a dual temperature acquisition mode can generate two virtual sensors from a single physical sensor. To quantify the components of the mixture from measurements on these (virtual) sensors, we have develop supervised and unsupervised source separation methods, applied to this nonlinear model: independent component analysis, least squares methods (Levenberg Marquardt algorithm), and a Bayesian method were studied. The experimental results show that these methods make it possible to estimate the VOC concentrations of a gas mixture, accurately, while requiring only a few calibration points.

Submitted by Maria Cristina (library@lncc.br) on 2016-07-27T18:52:15Z No. of bitstreams: 1 tese_Suelen.pdf: 922374 bytes, checksum: f324427616027a422decc0eaf56c7ae2 (MD5) / Approved for entry into archive by Maria Cristina (library@lncc.br) on 2016-07-27T18:52:32Z (GMT) No. of bitstreams: 1 tese_Suelen.pdf: 922374 bytes, checksum: f324427616027a422decc0eaf56c7ae2 (MD5) / Made available in DSpace on 2016-07-27T18:52:42Z (GMT). No. of bitstreams: 1 tese_Suelen.pdf: 922374 bytes, checksum: f324427616027a422decc0eaf56c7ae2 (MD5) Previous issue date: 2016-04-15 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Capes) / In this work a new method for obstacles reconstruction from partial boundary measurements is proposed. For a given boundary excitation, we want to determine the quantity, locations and sizes of a number of obstacles embedding whiting a geometrical domain, from partial boundary measurements related to such an excitation. This problem is written in the form of ill-posed and over-determinated partial differential equation. The idea therefore is to rewrite it as an optimization problem where a shape functional measuring the misfit between the boundary measurement and the solution to an auxiliary boundary value problem is minimized with respect to a set of ball-shaped holes. The topological derivative concept is used for solving the resulting topology optimization problem, leading to a second-order reconstruction algorithm free of initial guess. The resulting algorithm is non-iterative and thus very robust with respect to noisy data. Finally, some numerical results are presented in order to demonstrate the effectiveness of proposed reconstruction algorithm. / O objetivo deste trabalho é apresentar um novo método de reconstrução de obstáculos. Mais precisamente, dada uma excitação deseja-se obter a solução de um problema inverso de reconstrução consistindo na determinação da quantidade, localização e tamanho de obstáculos no interior de um dado domínio geométrico a partir de leituras parciais da resposta à referida excitação. Este problema é escrito na forma de uma equação diferencial parcial sobredeterminada. Essa dificuldade é contornada reescrevendo o problema inverso na forma de um problema de otimização. A ideia básica consiste em minimizar um funcional de forma que mede a diferença entre o dado lido e o calculado numericamente em relação ao próprio domínio geométrico. Em particular o conceito de derivada topológica é utilizado, o que conduz a um algoritmo de reconstrução de segunda ordem e independente de qualquer chute inicial. Como o algoritmo resultante é não-iterativo, o processo de reconstrução torna-se extremamente robusto à presença de ruído. Vários exemplos numéricos de reconstrução são apresentados donde se verifica a validade dos resultados obtidos.

Dans cette thèse, nous étudions un problème de contrôle et un problème inverse pour les équationsde la chaleur. Notre premier travail concerne la contrôlabilité à zéro pour une équation de la chaleur semi-linéaire. Il est à noter que sans contrôle, la solution est instable et il y aura en général explosion de la solution en un temps fini. Ici, nous proposons un résultat positif de contrôlabilité à zéro sous une hypothèse quantifiée de petitesse sur la donnée initiale. La nouveauté réside en la construction de ce contrôle pour amener la solution à l’état d’équilibre.Notre second travail aborde l’équation de la chaleur rétrograde dans un domaine borné et sous la condition de Dirichlet. Nous nous intéressons à la question suivante: peut-on reconstruire la donnée initiale à partir d’une observation de la solution restreinte à un sous-domaine et à un temps donné? Ce problème est connu pour être mal-posé. Ici, les deux principales méthodes proposées sont: une approche de filtrage des hautes fréquences et une minimisation à la Tikhonov. A chaque fois, nous reconstruisons de manière approchée la solution et quantifions l’erreur d’approximation / My thesis focuses on two main problems in studying the heat equation: Control problem and Inverseproblem.Our first concern is the null controllability of a semilinear heat equation which, if not controlled, can blow up infinite time. Roughly speaking, it consists in analyzing whether the solution of a semilinear heat equation, underthe Dirichlet boundary condition, can be driven to zero by means of a control applied on a subdomain in whichthe equation evolves. Under an assumption on the smallness of the initial data, such control function is builtup. The novelty of our method is computing the control function in a constructive way. Furthermore, anotherachievement of our method is providing a quantitative estimate for the smallness of the size of the initial datawith respect to the control time that ensures the null controllability property.Our second issue is the local backward problem for a linear heat equation. We study here the followingquestion: Can we recover the source of a linear heat equation, under the Dirichlet boundary condition, from theobservation on a subdomain at some time later? This inverse problem is well-known to be an ill-posed problem,i.e their solution (if exists) is unstable with respect to data perturbations. Here, we tackle this problem bytwo different regularization methods: The filtering method and The Tikhonov method. In both methods, thereconstruction formula of the approximate solution is explicitly given. Moreover, we also provide the errorestimate between the exact solution and the regularized one.

[pt] A introdução de um objeto estranho ao corpo humano pode resultar de diferentes processos que abrangem desde eventos iatrogênicos oriundos de procedimentos cirúrgicos, até eventos traumáticos, ocasionados por acidentes ou violência, como disparos de arma de fogo e perfurações com objetos cortantes. A presença desses corpos estranhos dentro do organismo humano pode ocasionar problemas de saúde que vão desde dor e incômodos ao óbito. Para remoção cirúrgica, torna-se essencial a localização desses objetos, com elevada exatidão, para redução do tempo cirúrgico e garantia de sucesso do procedimento. O presente trabalho tem por objetivo ampliar o estudo do problema de localização de corpos estranhos magnéticos (intrinsicamente magnético ou por indução magnética externa) no organismo humano. Foram desenvolvidos algoritmos computacionais capazes de promover simulações dos padrões de densidade de fluxo magnético em um plano decorrentes de uma fonte extensa, representada por parâmetros modificáveis de comprimento, posição espacial e inclinação ao plano de medição. Essas simulações foram utilizadas como entrada para o treinamento de Redes Neurais Artificiais que, após treinamento, foram capazes de resolver o problema inverso, caracterizando, a partir do mapa de campo magnético, a posição espacial, tamanho (comprimento) e inclinação da fonte metálica. Os resultados obtidos indicaram melhor desempenho com o uso de sensores triaxiais, cujo erro quadrático médio, em 3640 testes, mostrou-se inferior a 2 mm na orientação espacial, a 8 mm no comprimento e a 17 graus Celsius para a inclinação da fonte magnética em relação ao plano de medição. / [en] The introduction of a foreign object into the human body may result from different processes ranging from iatrogenic events during surgical procedures to traumatic events caused by accidents or violence such as firing and piercing with sharp objects. The presence of these foreign bodies within the human body can cause health problems ranging from pain and discomfort to death. For surgical removal, it is essential to locate these objects, with high accuracy, to reduce surgical time and guarantee the success of the procedure. The objective of the present work is to contribute to the study of the localization problem of magnetic foreign bodies (intrinsically magnetic or by external magnetic induction) in the human body. The developed computational algorithms are capable of promoting simulations of magnetic flux density patterns in a plane from an extensive source, represented by modifiable parameters of length, spatial position and slope to the measurement plane. These simulations were used as input for the training of Artificial Neural Networks that, after training, were able to solve the inverse problem, characterizing, from the magnetic field map, the spatial position, size (length) and slope of the metallic source. The results indicated a better performance with the use of triaxial sensors, whose mean square error, in 3640 tests, was less than 2 mm in the spatial orientation, 8 mm in length and 17 Celsius degrees for the tilt of the magnetic source in relation to the measurement plane.

Les facteurs de forme et les fonctions de corrélation déterminent les quantités dynamiques mesurables associées aux modèles de théorie des champs et de mécanique statistique. Dans le cas de modèles intégrables en dimension 2, au-delà des propriétés du spectre ou de la fonction de partition, un des grands défis actuels concerne le calcul exact des facteurs de forme et des fonctions de corrélation.Le but de cette thèse est de développer une approche permettant de résoudre ce problème dans le cadre de la méthode de séparation des variables quantique de Skyanin. Cette méthode généralise au cas quantique et pour des systèmes avec un grand nombre de degrés de liberté la méthode de Hamilton-Jacobi en mécanique analytique. Le Hamiltonien est exprimé avec des opérateurs séparés, son spectre et ses états propres caractérisés par un système d'équations de Baxter résultant des structures algébriques de Yang-Baxter, caractéristiques de l'intégrabilité de ces modèles.Cette thèse a permis, pour les modèles de sine-Gordon (théorie des champs quantique) et de Potts chiral (modèle de physique statistique), le calcul des produits scalaires entre états propres du Hamiltonien, la résolution du problème inverse, i. e. l'expression des opérateurs du modèle en termes des variables séparées, ainsi que le calcul en termes de déterminants des facteurs de forme, i. e. des éléments de matrice des opérateurs locaux du modèle dans la base propre du Hamiltonien, ce qui constitue un pas important vers le calcul des fonctions de corrélation de ces modèles. / Form factors and correlation functions determine the measurable dynamic quantities that are associated with field theories and statistical physics models. In the case of 2-dimensional integrable models, one of the main challenges beyond spectrum properties and partition function is the exact computation of form factors and correlation functions.The aim of this thesis is to develop an approach in the framework of Sklyanin's separation of variables to address this problem. This framework generalizes to the quantum case and for systems with many degrees of freedom the Hamilton-Jacobi method from analytical mechanics. The Hamiltonian is expressed in terms of separated operators, its spectrum and eigenvectors are characterized by a system of Baxter equations. These Baxter equations are a consequence of Yang-Baxter relations that are characteristic of these models being integrable.The result of this thesis is, in the case of the sine-Gordon model (quantum field theory) and of the chiral Potts model (statistical physics model), the computation of scalar products of Hamiltonian eigenstates, the resolution of the inverse problem (expressing the model operators in terms of separated variables) and the computation in terms of determinant of form factors (the matrix elements of the model local operators in the Hamiltonian eigenbasis), which is an important step towards the computation of the correlation functions of these models.

La cryo-microscopie électronique à particules isolées est une modalité d’imagerie permettant d’estimer la structure 3D de molécules. L’obtention d’un volume 3D est effectué par des algorithmes de reconstruction tomographique après acquisition par un microscope électronique à transmission d’un ensemble d’images de projection de l’objet observé. Les méthodes de reconstruction tomographique existantes permettent de déterminer la structure des molécules avec des résolutions proches de l’angström. Cependant la résolution est dégradée lorsque les molécules observées sont déformables. Les travaux réalisés au cours de cette thèse contribuent au développement de méthodes de traitement informatique des données (projections) dans le but de prendre en compte les déformations de l’objet observé dans le processus de reconstruction tomographique ab initio. Les problématiques principales abordées dans ce mémoire sont l’estimation des paramètres de projection basée sur la réduction de dimension non-linéaire, la détection des arêtes erronées dans les graphes de voisinages pour l’amélioration de la robustesse au bruit des méthodes de réduction de dimension, et la reconstruction tomographique basée sur un modèle paramétrique du volume. / Single particle cryo-electron microscopy is a technique that allows to estimate the 3D structure of biological complex. The construction of the 3D volume is performed by computerized tomography applied on a set of projection images from transmission electron microscope. Existing tomographic reconstructionalgorithms allow us to visualize molecular structure with a resolution around one angstrom. However the resolution is degraded when the molecules are deformable. This thesis contributes to the development of signal processing method in order to take into account the deformation information of the observed object for the ab initio tomographic reconstruction. The main contributions of this thesis are the estimation of projection parameters based on non-linear dimensionreduction, the false edges detection in neighborhood graphs to improve noise robustness of dimension reduction methods, and tomographic reconstruction based on a parametric model of the volume.