Since the origin of the automotive at the end of the 19th century, the traffic flow is subject to a constant increase and, unfortunately, involves a constant augmentation of road accidents. Research studies such as the one performed by the World Health Organization, show alarming results about the number of injuries and fatalities due to these accidents. To reduce these figures, a solution lies in the development of Advanced Driver Assistance Systems (ADAS) which purpose is to help the Driver in his driving task. This research topic has been shown to be very dynamic and productive during the last decades. Indeed, several systems such as Anti-lock Braking System (ABS), Electronic Stability Program (ESP), Adaptive Cruise Control (ACC), Parking Manoeuvre Assistant (PMA), Dynamic Bending Light (DBL), etc. are yet market available and their benefits are now recognized by most of the drivers. This first generation of ADAS are usually designed to perform a specific task in the Controller/Vehicle/Environment framework and thus requires only microscopic information, so requires sensors which are only giving local information about an element of the Vehicle or of its Environment. On the opposite, the next ADAS generation will have to consider more aspects, i.e. information and constraints about of the Vehicle and its Environment. Indeed, as they are designed to perform more complex tasks, they need a global view about the road context and the Vehicle configuration. For example, longitudinal control requires information about the road configuration (straight line, bend, etc.) and about the eventual presence of other road users (vehicles, trucks, etc.) to determine the best reference speed. [...]

Lors de ces travaux, nous nous sommes intéressés au problème de la classification supervisée d'images satellitaires de<br /> zones urbaines. Les données traitées sont des images optiques à très hautes résolutions spatiales: données panchromatiques à très haute résolution spatiale (IKONOS, QUICKBIRD, simulations PLEIADES) et des images hyperspectrales (DAIS, ROSIS).<br />Deux stratégies ont été proposées.<br />La première stratégie consiste en une phase d'extraction de caractéristiques spatiales et spectrales suivie d'une phase de classification. Ces caractéristiques sont extraites par filtrages morphologiques : ouvertures et fermetures géodésiques et filtrages surfaciques auto-complémentaires. La classification est réalisée avec les machines à vecteurs supports (SVM) <br /> non linéaires. Nous proposons la définition d'un noyau spatio-spectral utilisant de manière conjointe l'information spatiale<br /> et l'information spectrale extraites lors de la première phase.\\<br /> La seconde stratégie consiste en une phase de fusion de données pre- ou post-classification. Lors de la fusion postclassification,<br /> divers classifieurs sont appliqués, éventuellement sur plusieurs données issues d'une même scène (image panchromat<br />ique, image multi-spectrale). Pour chaque pixel, l'appartenance à chaque classe est estimée à l'aide des classifieurs. Un schém<br />a de fusion adaptatif permettant d'utiliser l'information sur la fiabilité locale de chaque classifieur, mais aussi l'information globale disponible a priori sur les performances de chaque algorithme pour les différentes classes, est proposé<br />.<br />Les différents résultats sont fusionnés à l'aide d'opérateurs flous.<br />Les méthodes ont été validées sur des images réelles. Des<br />améliorations significatives sont obtenues par rapport aux méthodes publiées dans la litterature.

The IEEE 802.11 networks are vulnerable to numerous wireless-specific attacks. Attackers can implement MAC address spoofing techniques to launch these attacks, while masquerading themselves behind a false MAC address. The implementation of Intrusion Detection Systems has become fundamental in the development of security infrastructures for wireless networks. This thesis proposes the designing a novel security system that makes use of metrics from multiple layers of observation to produce a collective decision on whether an attack is taking place. The Dempster-Shafer Theory of Evidence is the data fusion technique used to combine the evidences from the different layers. A novel, unsupervised and self- adaptive Basic Probability Assignment (BPA) approach able to automatically adapt its beliefs assignment to the current characteristics of the wireless network is proposed. This BPA approach is composed of three different and independent statistical techniques, which are capable to identify the presence of attacks in real time. Despite the lightweight processing requirements, the proposed security system produces outstanding detection results, generating high intrusion detection accuracy and very low number of false alarms. A thorough description of the generated results, for all the considered datasets is presented in this thesis. The effectiveness of the proposed system is evaluated using different types of injection attacks. Regarding one of these attacks, to the best of the author knowledge, the security system presented in this thesis is the first one able to efficiently identify the Airpwn attack.

Ces travaux de thèse portent sur la détection de défauts non francs dans des structures filaires particulières : les lignes de transmission a multiconducteurs (MTL), aussi appelées torons de câbles. Couramment employées pour le diagnostic de réseaux filaires, les méthodes par réflectométrie ne sont, pour l'heure, pas suffisamment performantes pour détecter de tels défauts. Par ailleurs, elles n'ont, en général, été étudiées et développées que pour des lignes simples, ou les phénomènes de couplages électromagnétiques entre les conducteurs (diaphonie) ne sont pas présents. Ces derniers sont cependant porteurs d'information supplémentaire sur l'état du câble. Les utiliser permettrait d'accroître la sensibilité de détection aux défauts. L'objectif est de proposer une nouvelle méthode de réflectométrie, tirant profit des signaux de diaphonie pour détecter les défauts non francs. Une telle méthode présente également l'avantage d'être adaptée aux structures en toron. Après avoir étudié l'impact d'un défaut non franc sur les paramètres caractéristiques d'une MTL et sur les signaux de diaphonie, une méthode, la "Cluster Time Frequency Domain Reflectometry ", a pu être proposée. Il s'agit d'un procédé en trois étapes. Des mesures par réflectométrie temporelle sont tout d'abord réalisées à l'entrée de la ligne à diagnostiquer. Tous les signaux présents, y compris ceux de diaphonie, sont enregistrés. Un traitement temps-fréquence leur est ensuite appliqué afin d'amplifier la présence d'éventuels défauts. Enfin, un algorithme de clustering, spécifiquement développé pour le diagnostic filaire, est utilisé de manière a bénéficier de l'ensemble de l'information disponible / Research works presented in this thesis rely on detecting soft faults (incipient faults) in specic wiring structures : multiconductor transmission lines (MTL), also known as bundles of wires. Reflectometry methods, often used for the diagnosis of wiring networks, aren't for now efficient enough at detecting such defects. Besides, they have been designed for single lines only, where electromagnetic coupling between conductors (crosstalk) is not to be considered. However such phenomenon can provide more information about the state of the cable. Using this information could enable us to detect soft faults more easily. Our goal is to propose a new reflectometry method, which takes advantage of crosstalk signals in order to detect incipient faults. Such a tool has also the advantage of being well-adapted to bundles of cables. Thanks to the preliminary study of the impact of soft faults on the characteristic parameters of a MTL and on crosstalk signals, a method called "Cluster Time Frequency Domain Reflectometry ", has been proposed. It is a three step process. Firts temporal reflectometry measurements are made at the beginning of the line under test. All the available signals, even crosstalk ones, are recorded. A time-frequency process is then applied on them, in order to amplify the presence of defects. Finally, a clustering algorithm, that has been specically developed for wiring diagnosis, is used to benefit from the whole available information

L'habitat intelligent est l'objet de nombreux travaux de recherche. Il permet d'assister des personnes âgées ou handicapées, d'améliorer le confort, la sécurité ou encore d'économiser de l'énergie. Aujourd'hui, l'informatique ubiquitaire se développe et s'intègre dans l'habitat intelligent notamment en apportant la sensibilité au contexte. Malheureusement, comprendre ce qui se passe dans une maison n'est pas toujours facile. Dans cette thèse, nous explicitons comment le contexte peut permettre de déployer des services adaptés aux activités et aux besoins des habitants. La compréhension du contexte passe par l'installation de capteurs mais aussi par l'abstraction des données brutes en données intelligibles facilement exploitables par des humains et des services. Nous mettons en avant une architecture multi-couches de fusion de données permettant d'obtenir des données contextuelles de niveaux d'abstraction différents. La mise en place des couches basses y est présentée en détail avec l'application de la théorie des fonctions de croyance pour l'abstraction de données brutes issues de capteurs. Enfin, sont présentés le déploiement d'un prototype nous ayant permis de valider notre approche, ainsi que les services déployés. / Smart home is a major subject of interest. It helps to assist elderly or disabled people, improve comfort, safety, and also save energy. Today, ubiquitous computing is developed and integrated into the smart home providing context-awareness. Unfortunately, understanding what happens in a home is not always easy. In this thesis, we explain how context can be used to deploy services tailored to the activities and needs of residents. Understanding context requires the installation of sensors but also the abstraction of raw data into easily understandable data usable by humans and services. We present a multi-layer architecture of data fusion used to obtain contextual information of different levels of abstraction. The implementation of the lower layers is presented in detail with the application of the theory of belief functions for the abstraction of raw sensor data. Finally, are presented the deployment of a prototype that allowed us to validate our approach and the deployed services.

Safe driving is a topic of multiple factors where the road surface condition is one. Knowledge about the road status can for instance indicate whether it is risk for low friction and thereby help increase the safety in traffic. The ambient temperature is an important factor when determining the road surface condition and is therefore in focus. This work evaluates different methods of data fusion to estimate the ambient temperature at road segments. Data from vehicles are used during the temperature estimation process while measurements from weather stations are used for evaluation. Both temporal and spatial dependencies are examined through different models to predict how the temperature will evolve over time. The proposed Kalman filters are able to both interpolate in road segments where many observations are available and to extrapolate to road segments with no or only a few observations. The results show that interpolation leads to an average error of 0.5 degrees during winter when the temperature varies around five degrees day to night. Furthermore, the average error increases to two degrees during springtime when the temperature instead varies about fifteen degrees per day. It is shown that the risk of large estimation error is high when there are no observations from vehicles. As a separate result, it has been noted that the weather stations have a bias compared to the measurements from the cars.

Dans le contexte du développement de systèmes embarqués alliant hautes performances et basse consommation, la recherche de l'efficacité énergétique optimale des processeurs est devenue un défi majeur. Les solutions architecturales se sont positionnées durant les dernières décennies comme d'importantes contributrices à ce challenge. Ces solutions, permettant la gestion du compromis performance de calcul/consommation, se sont dans un premier temps développées pour les circuits mono-processeurs. Elles évoluent aujourd'hui pour s'adapter aux contraintes de circuits MPSoCs de plus en plus complexes et sensibles aux déviations des procédés de fabrication, aux variations de tension et de température. Cette variabilité limite aujourd'hui drastiquement l'efficacité énergétique de chacune des unités de calcul qui composent une architecture MPSoC, car des marges pessimistes de fonctionnement sont généralement prises en compte. De grandes améliorations peuvent être attendues de la diminution de ces marges de fonctionnement en surveillant dynamiquement et localement la variabilité de chaque unité de calcul afin de réajuster ses paramètres de fonctionnement tension/fréquence. Ce travail s'insère dans une solution architecturale bas-coût nommée AVFS, basée sur une optimisation des techniques de gestion locales DVFS, permettant de réduire les marges de conception afin d'améliorer l'efficacité énergétique des MPSoCs, tout en minimisant l'impact de la solution proposée sur la surface de silicium et l'énergie consommée. Le développement d'un système de surveillance des variations locales et dynamiques de la tension et de la température à partir d'un capteur bas coût a été proposé. Une première méthode permet d'estimer conjointement la tension et la température à l'aide de tests statistiques. Une seconde permet d'accélérer l'estimation de la tension. Enfin, une méthode de calibration associée aux deux méthodes précédentes a été développée. Ce système de surveillance a été validé sur une plateforme matérielle afin d'en démontrer le caractère opérationnel. En prenant en compte les estimées de tension et de température, des politiques visant à réajuster dynamiquement les consignes des actionneurs locaux de tension et de fréquence ont été proposées. Finalement, la consommation additionnelle due à l'intégration des éléments constitutifs de l'architecture AVFS a été évaluée et comparée aux réductions de consommation atteignables grâce aux réductions des marges de fonctionnement. Ces résultats ont montré que la solution AVFS permet de réaliser des gains en consommation substantiels par rapport à une solution DVFS classique. / Nowadays, embedded systems requiring high performance and low power, the search for the optimal efficiency of the processors has become a major challenge. Architectural solutions have positioned themselves in recent decades as one of the main contributors to this challenge. These solutions enable the management of the trade-off between performance / power consumption, initially developed for single -processor systems. Today, they evolve to be adapted to the constraints of circuits MPSoCs increasingly complex and sensitive to process, voltage and temperature variations. This PVT variability limits drastically the energy efficiency of each of the processing units of a MPSoC architecture, taking into account pessimistic operating margins. Significant improvements can be expected from the reduction of the operating margins by dynamically monitoring and local variability of each resource and by adjusting its voltage / frequency operating point. This work is part of a low-cost architectural solution called AVFS, based on local DVFS optimization technique, to reduce design margins and improve the energy efficiency of MPSoCs, while minimizing the silicon surface and the energy additional cost. The development of a monitoring system of local and dynamic voltage and temperature variations using a low-cost sensor has been proposed. A first method estimates jointly voltage and temperature using statistical tests. A second one speeds up estimation of the voltage. Finally, a calibration method associated with the two previous methods has been developed. This monitoring system has been validated on a hardware platform to demonstrate its operational nature. Taking into account the estimation of voltage and temperature values, policies to dynamically adjust the set point of the local voltage and frequency actuators have been proposed. Finally, the additional power consumption due to the integration of the components of the architecture AVFS was evaluated and compared with reductions achievable through reductions in operating margins consumption. These results showed that the AVFS solution can achieve substantial power savings compared to conventional DVFS solution.

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.

[pt] Este trabalho estuda o problema de detecção de hipóteses binárias em sistemas distribuídos com centro de fusão operando em presença de canais seletivos em frequência. O uso de uma técnica de múltiplo acesso, referida aqui como CS-CDMA, é proposta para comunicação ortogonal entre os nós e o centro de fusão, assim como detector ótimo Bayesiano para fusão de dados em tais sistemas distribuídos é obtido. Como a complexidade do detector ótimo cresce exponencialmente com o número de nós sensores, um receptor sub-ótimo de baixa complexidade que realiza uma detecção casada multi-usuário seguida de decisão pela regra da maioria é proposto e examinado neste trabalho. Técnicas para estimação de canal, cega e assistida, necessárias para a implementação prática da detecção casada são também propostas. Simulações indicam que este receptor, de baixa complexidade, possui um desempenho próximo ao receptor ótimo. Com o objetivo de se ampliar o desempenho do detector casado do centro de fusão, é examinado o uso de cooperação na rede de sensores. Resultados de simulações mostraram que, como esperado, o uso de cooperação em sistema distribuídos utilizando o esquema de múltiplo acesso CS-CDMA melhora o desempenho do decisor do centro de fusão, entretanto esse ganho de desempenho mostrou-se mais significativo em ambientes com poucos multipercursos, uma vez que os sistemas distribuídos CS-CDMA não-cooperativos propostos exploram eficientemente a diversidade de multipercurso. Finalmente, este trabalho propõe um procedimento de fusão adaptativa não-assistida para sistemas distribuídos com fusão centralizada. Simulações mostram que a estratégia de fusão adaptativa possui desempenho muito próximo ao da regra de fusão ótima. / [en] This work studies the problem of detecting binary hypotheses in distributed systems with a fusion center operating in frequency selective channels. The use of a multiple access technique, referred herein as Chip Spread- Code Division Multiple Access (CS-CDMA), is proposed for orthogonal communication between the nodes and the fusion center and the Bayesian optimum detector for data fusion for such distributed systems is obtained. As the complexity of the optimal detector grows exponentially with the number of sensor nodes, a sub-optimal low-complexity receiver that performs a multi-user matched detection followed by the majority rule is proposed and examined in this work. Blind and assisted techniques for channel estimation necessary for the practical implementation of the matched detection have also been proposed. Simulations indicate that this low complexity receptor has a performance close to the optimal receiver. In order to increase the performance of the matched detector of the fusion center, it was examined the use of cooperation in this sensor network. Simulation results showed that, as expected, the use of cooperation in the distributed system with a multiple access scheme CS-CDMA improves the performance of the fusion center, however, this performance increasing was more significant in environments with few multipath, since the non-cooperative CS-CDMA distributed systems proposed here, efficiently exploits the multipath diversity. Finally, this paper proposes a non-assisted adaptive fusion for distributed systems with centralized fusion. Simulations show that the adaptive fusion strategy has a performance very close to the optimal fusion rule.

La fusion de données consiste à combiner plusieurs observations d'un environnement ou d'un phénomène afin de produire une description plus robuste, plus précise ou plus complète. Parmi les nombreux domaines d'application, les systèmes de surveillance multi capteurs étudiés dans ce travail occupent une place importante. Notre objectif est de fusionner les informations afin de compter le nombre de cibles, d'affiner la localisation et suivre les pistes en mouvement. D'un point de vue théorique, le problème a été abordé dans le contexte spécifique de la théorie des fonctions de croyance. Cette représentation qui constitue la première contribution originale de ce travail offre plusieurs avantages déterminants. Elle permet tout d'abord de modéliser des détections caractérisées par des incertitudes de géométries très différentes. Le modèle permet également d'intégrer des a priori topographiques en les modélisant par des BBAs spécifiques. Cette méthode d'intégration d'a priori constitue le deuxième élément orignal de ce travail. La troisième contribution concerne la définition d'un critère d'association entre les pistes et les détections à partir de la même représentation crédibiliste des localisations. Ce critère, maximisant la probabilité pignistique jointe des associations permet de réaliser de manière cohérente l'ensemble des traitements relatifs à la fusion sans avoir à définir un nouveau cadre de discernement. Malgré ces avantages, la taille du cadre de discernement exceptionnellement grande constitue un obstacle à l'exploitation de la théorie des croyances transférables. Pour contourner cette difficulté, chaque détection est projetée sur un cadre de discernement de plus petit cardinal grâce à une opération de conditionnement et de grossissement. De plus, le nombre d'éléments focaux peut augmenter considérablement en raison du caractère itératif de la fusion dans notre application. Afin de garder des temps de calcul raisonnables, il est donc impératif de simplifier régulièrement les BBAs. Ce point a fait l'objet d'une étude particulière à partir de laquelle une méthode de simplification reposant sur la décomposition canonique a été proposée. Enfin, au niveau système nous avons proposé une architecture décentralisée pour la réalisation de l'ensemble des traitements. Chaque nœud collabore alors avec ses voisins afin que les informations envoyées au poste de supervision forment un ensemble complet et cohérent. La validation du système de fusion a constitué une part importante de ce travail. Certains choix ont ainsi pu être justifiés en comparant les performances de différentes solutions envisageables au moyen de simulations. Parallèlement, la fusion a été testée lors de scénarios réels grâce à l'implantation d'un module dans le système de détection SmartMesh. Ces expériences ont été nécessaires d'une part pour quantifier de manière réaliste les erreurs relatives à chaque capteur mais aussi pour intégrer dans le plan de validation les difficultés liées aux interfaces avec les autres composants. / Data fusion combines several observations in order to produce a more accurate and complete description of the studied phenomenon. In this scope, the multi sensors detection system is a key element. In this work, we aim at merging information pieces from various sensors in order to count the objects or targets in the scene, localize and track the moving targets. In addition, when several targets are simultaneously present, we aim at managing multiple targets. In term of theoretic framework, the problem was addressed in the specific context of the belief functions theory. This choice implies the development of a credibilistic representation of the localization uncertainties such that each detection is modeled by a basic belief assignment (BBA) defined on a discrete paving of the scene. This model, which is the first contribution, allows us to represent the uncertainties about target location, respecting their specific geometric forms of imprecision, e.g. corresponding either to omnidirectional sensors or to directional sensors. As a second contribution, we propose to define a specific BBA, to represent and take into account some a priori knowledge such as topographic information pieces: obstacles partially occulting the vision or roads on which the target presence is more plausible. These BBAs are then merged with the detections to improve the localization. Based on the available belief model on target location, we also proposed (third contribution) a method for data association between tracks and detections. The new function to maximize is derived from the joint pignistic probability of associations. This criterion allows us to perform all processing in the same frame of discernment. Belief function theory main drawback for our application derives from the size of the frame of discernment. To address this issue, we have defined an adaptive discernment frame using conditioning and coarsening operators. Moreover, since the number of focal elements can also dramatically increase due to the iterative nature of our application, the BBAs should be regularly simplified. To address this issue, we have proposed a new method based on the canonical decomposition. Finally, the developed fusion system was implemented in a decentralized architecture. Each node cooperates with its neighbors to produce a coherent set of targets. Validation was the last but not least part of our work. Firstly, simulations were used to evaluate the proposed solutions versus the already existing methods. Secondly, the fusion process was tested in real-life scenarios by implementing a module in the SmartMesh surveillance system. These experiments were necessary both to quantify the errors for each sensor and to integrate difficulties related to interfaces with other components.