Thesis (Ph.D.) -- Mississippi State University. Department of Computer Science and Engineering. / Title from title screen. Includes bibliographical references.

Afin de respecter sa feuille de route, l’industrie du semi conducteur propose des nouvelles générations de technologies (appelées nœuds technologiques) tous les deux ans. Ces technologies présentent des dimensions de motifs de plus en plus réduites et par conséquent des contrôles des dimensions de plus en plus contraints. Cette réduction des tolérances sur les résultats métrologiques entraine forcément une évolution des outils de métrologie dimensionnelle. Aujourd’hui, pour les nœuds les plus avancés, aucune technique de métrologie ne peut répondre aux contraintes imposées. Les limitations se situent aussi bien sur les principes mêmes des méthodes employées que sur la quantité nécessaire de données permettant une analyse poussée ainsi que le temps de calcul nécessaire au traitement de ces données. Dans un contexte industriel, les aspects de rapidité et de précision des résultats de métrologie ne peuvent pas être négligés, de ce fait, une nouvelle approche fondée sur de la métrologie hybride doit être évaluée. La métrologie hybride consiste à mettre en commun différentes stratégies afin de combiner leurs forces et limiter leurs faiblesses. L’objectif d’une approche hybride est d’obtenir un résultat final présentant de meilleures caractéristiques que celui obtenu par chacune des techniques séparément. Cette problématique de métrologie hybride peut se résoudre par l’utilisation de la fusion de données. Il existe un grand nombre de méthodes de fusion de données couvrant des domaines très variés des sciences et qui utilisent des approches mathématiques différentes pour traiter le problème de fusion de données. L’objectif de ce travail de thèse est de développer cette problématique de métrologie hybride et fusion de données dans le cadre d’une collaboration entre deux laboratoires : LTM/ CNRS ( Laboratoire des Technologies de la Microélectronique) et le LETI/CEA (Laboratoire d’Electronique et de Technologies de l’Information). Le concept de la fusion de données est présenté dans un contexte de métrologie hybride appliquée au domaine de la microélectronique. L’état de l’art au niveau des techniques de métrologie est présenté et discuté. En premier lieu, le CD SEM pour ces caractéristiques associant rapidité et non destructibilité, ensuite l’AFM pour sa vision juste des profils des motifs et enfin la scattérométrie pour ses aspects de précision de mesures et sa rapidité tout en conservant une approche non destructive. Le FIB-STEM, bien que destructif, se positionne sur une approche de technique de référence. Les forces et les faiblesses de ces différentes méthodes sont évaluées afin de pouvoir les introduire dans une approche de métrologie hybride et d’identifier le rôle que chacune d’entre elle peut jouer dans ce contexte. Plusieurs campagnes de mesures ont été réalisées durant cette thèse afin d’apporter des connaissances sur les caractéristiques et les limitations de ces techniques et pouvoir les inclure dans différents scénarii de métrologie hybride. La méthode retenue pour la fusion de données est fondée sur une approche Bayesienne. Cette méthode a été évaluée dans un contexte expérimental cadré par un plan d’expérience permettant la mesure de la hauteur et la largeur de lignes en combinant différentes techniques de métrologie. Les données collectées ont été exploitées pour les étapes de debiaisage mais également pour un déroulement complet de fusion et dans les deux cas, la métrologie hybride montre les avantages de cette approche pour améliorer la justesse et la précision des résultats. Avec la poursuite d’un développement poussé, la technique de métrologie hybride présentée ici semble donc pouvoir s’intégrer dans un processus de fabrication dans l’industrie du semi conducteur. Son application n’est pas seulement destinée à de la métrologie dimensionnelle mais peut fournir également des informations sur la calibration des équipements. / The industry of semiconductors continues to evolve at a fast pace, proposing a new technology node around every two years. Each new technology node presents reduced feature sizes and stricter dimension control. As the features of devices continue to shrink, allowed tolerances for metrology errors must shrink as well, pushing the evolution of the metrology tools.No individual metrology technique alone can answer the tight requirements of the industry today, not to mention in the next technology generations. Besides the limitations of the metrology methods, other constraints such as the amount of metrology data available for higher order analysis and the time required for generating such data are also relevant and impact the usage of metrology in production. For the production of advanced technology nodes, neither speed nor precision may be sacrificed, which calls for cleverer metrology approaches, such as the Hybrid Metrology.Hybrid Metrology consists of employing different metrology strategies together in order to combine their strengths while mitigating their weaknesses. This hybrid approach goal is to improve the measurements in such a way that the final data presents better characteristics that each method separately. One of the techniques than can be used to combine the data coming from different metrology techniques is called Data Fusion. There are a large number of developed methods of Data Fusion, using different mathematical tools, to address the data fusion process.The first goal of this thesis project was to start developing the topics of Data Fusion and Hybrid Metrology within the two laboratories whose cooperation made this work possible: LTM (Laboratoire des Technologies de la Microélectronique) and LETI (Laboratoire d'électronique et de technologie de l'information). This thesis presents the concepts of Data Fusion in the context of Hybrid Metrology applied to dimensional measuring for the semiconductors industry. This concept can be extensively used in many other fields of applications.In this work the basics of state-of-the-art metrology techniques is presented and discussed. The focus is the CD-SEM, for its fast and almost-non-destructive metrology; the AFM, for its accurate profile view of patterns and non-destructive characteristic; the Scatterometry, for its precision, global and fast measurements; and the FIB-STEM, as a reference on accuracy for any type of profile, although destructive. The strengths and weaknesses of these methods were discussed in order to introduce the need of Hybrid Metrology and to identify the role that each of those methods can play in this context.Several experiments were performed during this thesis work in order to provide further knowledge about the characteristics and limitations of each metrology method and to be used as either inputs or reference on the different Hybrid Metrology scenarios proposed.The selected method for fuse the data coming from different metrology methods was the Bayesian approach. This technique was evaluated in different experimental contexts, both for Height and CD metrology combining different metrology methods. Results were evaluated for both the debiasing step alone and for the complete fusion flow. In both cases, it was clear the advantages of using a Hybrid Metrology approach for improving the measurement precision and accuracy.The presented Hybrid Metrology technique may be used by the semiconductor industry in different steps of the fabrication process. This technique can also provide information for machine calibration, such as a CD-SEM tool being calibrated based on Hybrid Metrology results generated using the CD-SEM itself together with Scatterometry data.

In a dynamically monitored environment the analyst team need timely and accurate information to conduct proactive action over complex situations. Typically, there are thousands of reported activities in a real time operation, therefore steps are taken to direct the analyst’s attention to the most important activity. The data fusion community have introduced the information fusion model, with multiple situational assessments. Each process lends itself to ranking the most important activities into a predetermined order. Unfortunately, the capability of a real time system can be hindered by the knowledge limitation problem, particularly when the underlying system is processing multiple sensor information. Consequently, the situational awareness domains may not rank the identified situation as perfect, as desired by the decision-making resources. This thesis presents advanced research carried out to evaluate the ranking capability of information from the situational awareness domains: perception, comprehension and projection. The Ranking Capability Score (RCS) has been designed for evaluating the prioritisation process. The enhanced (RCS) has been designed for addressing the knowledge representation problem in the user system relation under a situational assessment where the proposed number of tracking activities are dynamically shifted. Finally, the Scheduling Capability Score was designed for evaluating the scheduling capability of the situational awareness system. The proposed performance metrics have been successful in fulfilling their objectives. Furthermore, they have been validated and evaluated using an analytical approach, through conducting a rigorous analysis of the prioritisation and scheduling processes, despite any constraints related to a domain-specific configuration.

L'objectif de cette thèse est d'améliorer ou de préserver les performances d'un système décisionnel quand l’environnement peut impacter certains attributs de l'espace de représentation à un instant donné ou en fonction de la position géographique de l’observation. S'inspirant des méthodes d'ensemble, notre approche a consisté à prendre les décisions dans des sous-espaces de représentation résultant de projections de l'espace initial, espérant ainsi travailler dans des sous-espaces non impactés. La décision finale est alors prise par fusion des décisions individuelles. Dans ce contexte, trois méthodes de classification (one-class SVM, Kernel PCA et Kernel ECA) ont été testées en segmentation d'images texturées qui constitue un support applicatif parfaitement adéquat en raison des ruptures de modèle de texture aux frontières entre deux régions. Ensuite, nous avons proposé une nouvelle règle de fusion reposant sur un test du rapport de vraisemblance pour un ensemble de classifieurs indépendants. Par rapport au vote majoritaire, cette règle de fusion a montré de meilleures performances face à l'altération de l'espace de représentation. Enfin, nous avons établi un modèle conjoint pour l’ensemble des variables décisionnelles de Bernoulli corrélées associées aux décisions des classifieurs individuels. Cette modélisation doit permettre de lier les performances des classifieurs individuels à la performance de la règle de décision globale et d’étudier et de maîtriser l'impact des changements de l'espace initial sur la performance globale / The objective of this thesis is to improve or maintain the performance of a decision-making system when the environment can impact some attributes of the feature space at a given time or depending on the geographical location of the observation. Inspired by ensemble methods, our approach has been to make decisions in representation sub-spaces resulting of projections of the initial space, expecting that most of the subspaces are not impacted. The final decision is then made by fusing the individual decisions. In this context, three classification methods (one-class SVM, Kernel PCA and Kernel ECA) were tested on a textured images segmentation problem which is a perfectly adequate application support because of texture pattern changes at the border between two regions. Then, we proposed a new fusion rule based on a likelihood ratio test for a set of independent classifiers. Compared to the majority vote, this fusion rule showed better performance against the alteration of the performance space. Finally, we modeled the decision system using a joint model for all decisions based on the assumption that decisions of individual classifiers follow a correlated Bernoulli law. This model is intended to link the performance of individual classifiers to the performance of the overall decision rule and to investigate and control the impact of changes in the original space on the overall performance

O uso de veículos aéreos não tripulados (VANTs) tem se tornado cada vez mais comum, principalmente em aplicações de uso civil. No cenário militar, o uso de VANTs tem focado o cumprimento de missões específicas que podem ser divididas em duas grandes categorias: sensoriamento remoto e transporte de material de emprego militar. Este trabalho se concentra na categoria do sensoriamento remoto. O trabalho foca a definição de um modelo e uma arquitetura de referência para o desenvolvimento de sensores inteligentes orientados a missões específicas. O principal objetivo destas missões é a geração de mapas temáticos. Neste trabalho são investigados processos e mecanismos que possibilitem a geração desta categoria de mapas. Neste sentido, o conceito de MOSA (Mission Oriented Sensor Array) é proposto e modelado. Como estudos de caso dos conceitos apresentados são propostos dois sistemas de mapeamento automático de fontes sonoras, um para o caso civil e outro para o caso militar. Essas fontes podem ter origem no ruído gerado por grandes animais (inclusive humanos), por motores de combustão interna de veículos ou por atividade de artilharia (incluindo caçadores). Os MOSAs modelados para esta aplicação são baseados na integração de dados provenientes de um sensor de imageamento termal e uma rede de sensores acústicos em solo. A integração das informações de posicionamento providas pelos sensores utilizados, em uma base cartográfica única, é um dos aspectos importantes tratados neste trabalho. As principais contribuições do trabalho são a proposta de sistemas MOSA, incluindo conceitos, modelos, arquitetura e a implementação de referência representada pelo sistema de mapeamento automático de fontes sonoras. / The use of unmanned aerial vehicles (UAV) has become increasingly common, particularly in civilian applications. In the military scenario, the use of UAVs has focused the accomplishment of specific tasks in two broad categories: remote sensing and transport of military material. This work focuses the remote sensing category. It address the definition of a model and reference architecture for the development of smart sensors oriented to specific tasks. The main objective of these missions is to generate thematic maps. This work investigates processes and mechanisms that enable the automatic generation of thematic maps. In this sense, the concept of MOSA (Mission Oriented Sensor Array) is proposed and modeled. As case studies, we propose two automatic mapping systems for on-the-ground generated sound sources, one for the civilian case and one for the military case. These sounds may come from the noise generated by large animals (including humans), from internal combustion engine vehicles or from artillery activity (including hunters). The MOSAs modeled for this application integrate data from a thermal imaging sensor and an on-the-ground network of acoustic sensors. The fusion of position information, provided by the two sensors, into a single cartographic basis is one of the key aspects addressed in this work. The main contributions are the proposed MOSA systems, including concepts, models and architecture and the reference implementation comprised by the system for automatic mapping of sound sources.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2016-02-01T07:36:16Z No. of bitstreams: 2 Tese - Mariana da Silva Godinho - 2014.pdf: 11659350 bytes, checksum: 36a0ea84b6d0c992d5629322e89390a3 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2016-02-01T07:42:02Z (GMT) No. of bitstreams: 2 Tese - Mariana da Silva Godinho - 2014.pdf: 11659350 bytes, checksum: 36a0ea84b6d0c992d5629322e89390a3 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2016-02-01T07:42:02Z (GMT). No. of bitstreams: 2 Tese - Mariana da Silva Godinho - 2014.pdf: 11659350 bytes, checksum: 36a0ea84b6d0c992d5629322e89390a3 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2014-06-27 / Conselho Nacional de Pesquisa e Desenvolvimento Científico e Tecnológico - CNPq / The lifetime of a power transformer is determined by the state of degradation of its insulation system. This system consists of Kraft paper immersed in insulating oil (IMO). In order to prevent an abrupt crash transformer its oxidation state should be monitored. This monitoring is performed using physico-chemical and chromatographic assays. The physico-chemical parameter of greatest importance is the interfacial tension (IT), which is a measure for assessing indirectly the oxidation state of the Kraft paper. The objective of this thesis to propose analytical methodologies to evaluate the oxidation process of the insulation system, both in naturally and accelerated degraded oil samples, using Image Analysis and spectroscopic techniques combined with methods of Multivariate Calibration (Partial Least Squares - PLS), Resolution Multivariate Curve and Data Fusion. Spectra of Moleculular Fluorescence (MF), Nuclear Magnetic Resonance of 1H (1H NMR), Fourier Transform Near Infrared (FT-NIR) and RGB images were obtained for a set of 268 oil samples provided by CELG Distribuição S.A. (CELG D). Subsequently, spectra in the Ultraviolet-Visible (UV-Vis) region, FM and RGB images were obtained for a set of 24 oil samples subjected to an accelerated degradation. PLS models were obtained in order to predict the IT and color parameters for the set of samples provided by CELG D. For the set of degraded samples PLS models were obtained for prediction of degradation time. PARAFAC and MCR-ALS models were obtained to evaluate the oxidation process of the insulating system and too identify the pure spectra of possible compounds formed during this process. Furthermore, low level data fusion combined with variable selection using the Variables Importance in the Projection (VIPs) was used in this study to obtain PLS models. The best PLS model for the set of samples provided by CELG D was obtained for the fused spectroscopic data using variable selection by VIPS. This model provided a mean square error of prediction (RMSEP) of 2.9 mN/m for IT and 0.3 for the color parameter (dimensionless units). PARAFAC and MCR-ALS models allowed to assess the process of system degradation. The proposed methodologies are fast and non-destructive, showing great advantages over traditional methods, which take a long time, and generate a lot of waste. This work opens perspectives for an online monitoring to assess the degradation of the insulating system besides having permitted a more detailed study on the compounds formed during the degradation process. / A vida útil de um transformador de potência é determinada pelo estado de degradação de seu sistema isolante. Esse sistema é constituído pelo papel tipo Kraft imerso em óleo mineral isolante (OMI). Para evitar que um transformador chegue ao seu fim de vida de forma abrupta é necessário que seja realizado o monitoramento do seu estado de oxidação. Esse monitoramento é feito através de ensaios físico-químicos e cromatográficos. O parâmetro físico-químico de maior importância é a tensão interfacial (TI), pois é uma medida que avalia indiretamente o estado de oxidação do papel. O objetivo deste trabalho foi propor metodologias analíticas para avaliar o processo de degradação do sistema isolante tanto em amostras degradadas naturalmente quanto em amostras degradadas aceleradamente, usando Análise de Imagens e técnicas espectroscópicas aliadas a métodos de Calibração Multivariada (Mínimos Quadrados Parciais - PLS), Resolução de Curva Multivariada e Fusão de Dados. Foram obtidos espectros de Fluorescência Molecular (FM), Ressonância Magnética Nuclear de 1H (RMN de 1H), Infravermelho na região do próximo (FT-NIR) e imagens RGB de um conjunto de 268 amostras de OMI fornecidas pela CELG Distribuição S.A. (CELG D). Posteriormente, foram obtidos espectros de UV-Vis, FM e imagens RGB de um conjunto de 24 amostras degradadas aceleradamente. Modelos PLS foram construídos para prever os parâmetros TI e cor para o conjunto de amostras fornecidas pela CELG D. Para as amostras degradadas aceleradamente modelos PLS foram obtidos para previsão do tempo de degradação. Modelos PARAFAC e MCR-ALS foram construídos com objetivo de avaliar o processo de oxidação do sistema isolante e identificar os espectros puros de possíveis compostos formados durante esse processo. Além disso, Fusão de Dados de baixo nível aliada à seleção de variáveis usando as Variáveis de Importância na Projeção (VIPs) foram utilizadas neste trabalho para a obtenção de modelos PLS. O melhor modelo PLS para o conjunto de amostras fornecidas pela CELG D foi obtido para os dados espectroscópicos fundidos usando seleção de variáveis por VIPS. Esse modelo forneceu um erro médio quadrático de previsão (RMSEP) de 2,9 mN/m para previsão da TI e 0,3 para o parâmetro cor (unidade adimensional). Os modelos PARAFAC e MCR-ALS permitiram avaliar o processo de degradação do sistema isolante. As metodologias propostas são rápidas e não destrutivas, mostrando grandes vantagens sobre os métodos tradicionais, os quais demandam muito tempo, além de gerar uma grande quantidade de resíduos. Este trabalho abre perspectivas para um monitoramento direto (on-line) para avaliar a degradação do sistema isolante, além de ter permitido um estudo mais aprofundado a respeito dos compostos formados durante o processo de degradação.

Fruit d’un travail collaboratif entre le LAAS-CNRS de Toulouse et l’entreprise MEAS-France / TE Connectivity, ces travaux ont consisté en la mise en place d’une méthodologie permettant la réalisation de capteurs embarqués intelligents utilisant la fusion de données multi-physique pour estimer un paramètre en amoindrissant l’impact des variations environnementales.Nous explorons ici les méthodes liées à la modélisation et l’estimation de paramètres au travers des filtres de Kalman, pour les systèmes linéaires, et des filtres de Kalman étendus (EKF) et Unscented Kalman Filter pour les systèmes non-linéaires. Nous proposons ensuite des méthodes hybrides permettant d’obtenir le meilleur rapport charge de calculs/précision pour les systèmes présentant une évolution linéaire et une mesure non-linéaire.Après une étude de la complexité algorithmique des différentes solutions, nous proposons des méthodes permettant de diminuer la charge de calculs afin de satisfaire les contraintes temps-réel avec une faible puissance de calculs, telles que trouvées couramment dans les applications embarquées. La méthode développée est finalement appliquée sur deux cas applicatifs concrets : le capteur de qualité d’urée de la société MEAS-France/TE Connectivity et le capteur d’analyse du mouvement AREM développés au cours de la thèse au sein du LAAS-CNRS. / The work detailed in this document is the result of a collaborative effort of the LAAS-CNRS in Toulouse and MEAS-France / TE Connectivity during a period of three years.The goal here is to develop a methodology to design smart embedded sensors with the ability to estimate physical parameters based on multi-physical data fusion. This strategy tends to integrate sensors technologies, currently dedicated to lab measurements, in low powered embedded systems working in imperfects environments. After exploring model oriented methods, parameters estimations and Kalman filters, we detail various existing solutions upon which we can build a valid response to multi-physical data fusion problematics, for linear systems with the Kalman Filter, and for non-linear systems with the Extended Kalman Filter and the Unscented Kalman Filter.Then, we will synthesize a filter for hybrid systems, having a linear evolution model and a non-linear measurement model. For example, using the best of the two worlds in order to obtain the best complexity/precision ratio. Once we selected the estimation method, we detail computing power and algorithm complexity problematics in order to find available optimizations we can use to assess the usability of our system in a low power environment. Then we present the developed methodology application to the UQS sensor, sold by TE Connectivity, study case. This sensor uses near infra-red spectroscopy to determine the urea concentration in a urea/water solution, in order to control the nitrogen-oxyde depolluting process in gasoline engines. After a design principles presentation, we detail the model we created in order to represent the system, to simulate its behavior and to combine the measurement data to extract the desired concentration. During this step, we focus on the obstacles of our model calibration and the deviation compensation, due toworking conditions or to components aging process. Based on this development, we finally designed the hybrid models addressing the nominal working cases and the model re-calibration during the working duration of the product. After this, we presented obtained results, on simulated data, and on real-world measured data. Finally, we enhanced the methodology based on tabulated “black box” models which are easier to calibrate and cheaper to process. In conclusion, we reapplied our methodology to a different motion capture sensor, to compile all possible solutions and limits.

Le thème de recherche de cette thèse est la perception coopérative multi-véhicules appliquée au contexte des systèmes de véhicules intelligents. L’objectif général des travaux présentés dans cette thèse est de réaliser la perception coopérative de plusieurs véhicules (dite « perception coopérative »), visant ainsi à fournir des résultats de perception améliorés par rapport à la perception d’un seul véhicule (ou « perception non-coopérative »). Au lieu de concentrer nos recherches sur la performance absolue de la perception coopérative, nous nous concentrons sur les mécanismes généraux qui permettent la réalisation de la localisation coopérative et de la cartographie de l’environnement routier (y compris la détection des objets), considérant que la localisation et la cartographie sont les deux tâches les plus fondamentales pour un système de véhicule intelligent. Nous avons également exploité la possibilité d’explorer les techniques de la réalité augmentée, combinées aux fonctionnalités de perception coopérative. Nous baptisons alors cette approche « réalité augmentée coopérative ». Par conséquent, nous pouvons d’ores et déjà annoncer trois contributions des travaux présentés: la localisation coopérative, la cartographie locale coopérative, et la réalité augmentée coopérative. / The research theme of this dissertation is the multiple-vehicles cooperative perception (or cooperative perception) applied in the context of intelligent vehicle systems. The general methodology of the presented works in this dissertation is to realize multiple-intelligent vehicles cooperative perception, which aims at providing better vehicle perception result compared with single vehicle perception (or non-cooperative perception). Instead of focusing our research works on the absolute performance of cooperative perception, we focus on the general mechanisms which enable the realization of cooperative localization and cooperative mapping (and moving objects detection), considering that localization and mapping are two underlying tasks for an intelligent vehicle system. We also exploit the possibility to realize certain augmented reality effect with the help of basic cooperative perception functionalities; we name this kind of practice as cooperative augmented reality. Naturally, the contributions of the presented works consist in three aspects: cooperative localization, cooperative local mapping and moving objects detection, and cooperative augmented reality.

La fabrication de pièces manufacturées génère un nombre très important de données de différents types définissant les géométries de fabrication ainsi que la qualité de production. Ce travail de thèse s’inscrit dans le cadre de la réalisation d’un système de vision cognitif dédié à l’évaluation d’objets 3D manufacturés incluant éventuellement des surfaces gauches, en tenant compte des tolérances géométriques et des incertitudes. Ce système permet un contrôle exhaustif de pièces manufacturées et offre la possibilité d’une inspection tridimensionnelle automatique de la pièce. La mise en place d’un système de mesures multi-capteurs (passifs et actifs) a permis d’améliorer significativement la qualité d’évaluation par le biais d’une reconstruction tridimensionnelle enrichie de l’objet à évaluer. En particulier, nous avons employé simultanément un système stéréoscopique de vision et un système à projection de lumière structurée afin de reconstruire les contours et les surfaces de différents objets 3D. / Producing industrially manufactured parts generates a very large number of data of various types defining the manufacturing geometries as well as the quality of production. This PhD work has been carried out within the framework of the realization of a cognitive vision system dedicated to the 3D evaluation of manufactured objects including possibly free form surfaces, taking into account the geometric tolerances and uncertainties. This system allows the comprehensive control of manufactured parts, and provides the means for their automated 3D dimensional inspection. The implementation of a multi-sensor (passive and active) measuring system enabled to improve significantly the assessment quality through an enriched three-dimensional reconstruction of the object to be evaluated. Specifically, we made use simultaneously of a stereoscopic vision system and of a structured light based system in order to reconstruct the edges and surfaces of various 3D objects.

Le tri automatique des déchets est un sujetcomplexe en raison de la diversité des objets et desmatériaux présents. Il nécessite un apport de donnéesvariées et hétérogènes. Cette thèse traite du problème defusion de données découlant d’un dispositif de troiscapteurs dont une caméra hyperspectrale dans ledomaine NIR. Nous avons étudié l’avantage d’utiliser lecadre des fonctions de croyance (BFT) tout au long de ladémarche de fusion en utilisant notamment la mesure deconflit comme un critère clé de notre approche. Dans unepremière partie, nous avons étudié l'intérêt de la BFTpour la classification multiclasse des donnéeshyperspectrales à partir d’Error Correcting OutputCodes (ECOC) qui consistent à séparer le problèmemulticlasse en un ensemble de sous-problèmes binairesplus simples à résoudre. Les questions de commentidéalement séparer le problème multiclasse (codage)ainsi que celle de la combinaison des réponses de cesproblèmes binaires (décodage) sont encore aujourd’huides questions ouvertes. Le cadre des fonctions decroyance permet de proposer une étape de décodage quimodélise chaque classifieur binaire comme une sourceindividuelle d'information grâce notamment à lamanipulation des hypothèses composées. Par ailleurs laBFT fournit des indices pour détecter les décisions peufiables ce qui permet une auto-évaluation de la méthoderéalisée sans vérité terrain. Dans une deuxième partietraitant de la fusion de données, nous proposons unedémarche ‘orientée-objet’ composée d’un module desegmentation et d’un module de classification afin defaire face aux problèmes d’échelle, de différences derésolutions et de recalage des capteurs. L’objectif estalors d’estimer une segmentation où les segmentscoïncident avec les objets individuels et sont labellisés entermes de matériau. Nous proposons une interactionentre les modules à base de validation mutuelle. Ainsi,d’une part la fiabilité de la labellisation est évaluée auniveau des segments, d’autre part l’information declassification interagit sur les segments initiaux pour serapprocher d’une segmentation au niveau « objet » : leconsensus (ou l’absence de consensus) parmi lesinformations de classification au sein d’un segment ouentre segments connexes permet de faire évoluer lesupport spatial vers le niveau objet. / Automatic waste sorting is a complex matterbecause of the diversity of the objects and of the presentmaterials. It requires input from various andheterogeneous data. This PhD work deals with the datafusion problem derived from an acquisition devicecomposed of three sensors, including an hyperspectralsensor in the NIR field. We first studied the benefit ofusing the belief function theory framework (BFT)throughout the fusion approach, using in particularconflict measures to drive the process. We first studiedthe BFT in the multiclass classification problem createdby hyperspectral data. We used the Error CorrectingOutput Codes (ECOC) framework which consists inseparating the multiclass problem into several binaryones, simpler to solve. The questions of the idealdecomposition of the multiclass problem (coding) and ofthe answer combination coming from the binaryclassifiers (decoding) are still open-ended questions. Thebelief function framework allows us to propose adecoding step modelling each binary classifier as anindividual source of information, thanks to the possibilityof handling compound hypotheses. Besides, the BFTprovides indices to detect non reliable decisions whichallow for an auto-evaluation of the method performedwithout using any ground truth. In a second part dealingwith the data fusion,we propose an evidential version ofan object-based approach composed with a segmentationmodule and a classification module in order to tackle theproblems of the differences in scale, resolutions orregistrations of the sensors. The objective is then toestimate a relevant spatial support corresponding to theobjects while labelling them in terms of material. Weproposed an interactive approach with cooperationbetween the two modules in a cross-validation kind ofway. This way, the reliability of the labelling isevaluated at the segment level, while the classificationinformation acts on the initial segments in order toevolve towards an object level segmentation: consensusamong the classification information within a segment orbetween adjacent regions allow the spatial support toprogressively reach object level