Reconocimiento facial

El presente proyecto trata sobre uno de los campos más problemáticos de la inteligencia artificial, el reconocimiento facial. Algo tan sencillo para las personas como es reconocer una cara conocida se traduce en complejos algoritmos y miles de datos procesados en cuestión de segundos. El proyecto comienza con un estudio del estado del arte de las diversas técnicas de reconocimiento facial, desde las más utilizadas y probadas como el PCA y el LDA, hasta técnicas experimentales que utilizan imágenes térmicas en lugar de las clásicas con luz visible. A continuación, se ha implementado una aplicación en lenguaje C++ que sea capaz de reconocer a personas almacenadas en su base de datos leyendo directamente imágenes desde una webcam. Para realizar la aplicación, se ha utilizado una de las librerías más extendidas en cuanto a procesado de imágenes y visión artificial, OpenCV. Como IDE se ha escogido Visual Studio 2010, que cuenta con una versión gratuita para estudiantes. La técnica escogida para implementar la aplicación es la del PCA ya que es una técnica básica en el reconocimiento facial, y además sirve de base para soluciones mucho más complejas. Se han estudiado los fundamentos matemáticos de la técnica para entender cómo procesa la información y en qué se datos se basa para realizar el reconocimiento. Por último, se ha implementado un algoritmo de testeo para poder conocer la fiabilidad de la aplicación con varias bases de datos de imágenes faciales. De esta forma, se puede comprobar los puntos fuertes y débiles del PCA. ABSTRACT. This project deals with one of the most problematic areas of artificial intelligence, facial recognition. Something so simple for human as to recognize a familiar face becomes into complex algorithms and thousands of data processed in seconds. The project begins with a study of the state of the art of various face recognition techniques, from the most used and tested as PCA and LDA, to experimental techniques that use thermal images instead of the classic visible light images. Next, an application has been implemented in C + + language that is able to recognize people stored in a database reading images directly from a webcam. To make the application, it has used one of the most outstretched libraries in terms of image processing and computer vision, OpenCV. Visual Studio 2010 has been chosen as the IDE, which has a free student version. The technique chosen to implement the software is the PCA because it is a basic technique in face recognition, and also provides a basis for more complex solutions. The mathematical foundations of the technique have been studied to understand how it processes the information and which data are used to do the recognition. Finally, an algorithm for testing has been implemented to know the reliability of the application with multiple databases of facial images. In this way, the strengths and weaknesses of the PCA can be checked.

Multimodal integration of functional and structural brain connectivity. An application to the characterization of Mild Cognitive Impairment

La investigación para el conocimiento del cerebro es una ciencia joven, su inicio se remonta a Santiago Ramón y Cajal en 1888. Desde esta fecha a nuestro tiempo la neurociencia ha avanzado mucho en el desarrollo de técnicas que permiten su estudio. Desde la neurociencia cognitiva hoy se explican muchos modelos que nos permiten acercar a nuestro entendimiento a capacidades cognitivas complejas. Aun así hablamos de una ciencia casi en pañales que tiene un lago recorrido por delante. Una de las claves del éxito en los estudios de la función cerebral ha sido convertirse en una disciplina que combina conocimientos de diversas áreas: de la física, de las matemáticas, de la estadística y de la psicología. Esta es la razón por la que a lo largo de este trabajo se entremezclan conceptos de diferentes campos con el objetivo de avanzar en el conocimiento de un tema tan complejo como el que nos ocupa: el entendimiento de la mente humana. Concretamente, esta tesis ha estado dirigida a la integración multimodal de la magnetoencefalografía (MEG) y la resonancia magnética ponderada en difusión (dMRI). Estas técnicas son sensibles, respectivamente, a los campos magnéticos emitidos por las corrientes neuronales, y a la microestructura de la materia blanca cerebral. A lo largo de este trabajo hemos visto que la combinación de estas técnicas permiten descubrir sinergias estructurofuncionales en el procesamiento de la información en el cerebro sano y en el curso de patologías neurológicas. Más específicamente en este trabajo se ha estudiado la relación entre la conectividad funcional y estructural y en cómo fusionarlas. Para ello, se ha cuantificado la conectividad funcional mediante el estudio de la sincronización de fase o la correlación de amplitudes entre series temporales, de esta forma se ha conseguido un índice que mide la similitud entre grupos neuronales o regiones cerebrales. Adicionalmente, la cuantificación de la conectividad estructural a partir de imágenes de resonancia magnética ponderadas en difusión, ha permitido hallar índices de la integridad de materia blanca o de la fuerza de las conexiones estructurales entre regiones. Estas medidas fueron combinadas en los capítulos 3, 4 y 5 de este trabajo siguiendo tres aproximaciones que iban desde el nivel más bajo al más alto de integración. Finalmente se utilizó la información fusionada de MEG y dMRI para la caracterización de grupos de sujetos con deterioro cognitivo leve, la detección de esta patología resulta relevante en la identificación precoz de la enfermedad de Alzheimer. Esta tesis está dividida en seis capítulos. En el capítulos 1 se establece un contexto para la introducción de la connectómica dentro de los campos de la neuroimagen y la neurociencia. Posteriormente en este capítulo se describen los objetivos de la tesis, y los objetivos específicos de cada una de las publicaciones científicas que resultaron de este trabajo. En el capítulo 2 se describen los métodos para cada técnica que fue empleada: conectividad estructural, conectividad funcional en resting state, redes cerebrales complejas y teoría de grafos y finalmente se describe la condición de deterioro cognitivo leve y el estado actual en la búsqueda de nuevos biomarcadores diagnósticos. En los capítulos 3, 4 y 5 se han incluido los artículos científicos que fueron producidos a lo largo de esta tesis. Estos han sido incluidos en el formato de la revista en que fueron publicados, estando divididos en introducción, materiales y métodos, resultados y discusión. Todos los métodos que fueron empleados en los artículos están descritos en el capítulo 2 de la tesis. Finalmente, en el capítulo 6 se concluyen los resultados generales de la tesis y se discuten de forma específica los resultados de cada artículo. ABSTRACT In this thesis I apply concepts from mathematics, physics and statistics to the neurosciences. This field benefits from the collaborative work of multidisciplinary teams where physicians, psychologists, engineers and other specialists fight for a common well: the understanding of the brain. Research on this field is still in its early years, being its birth attributed to the neuronal theory of Santiago Ramo´n y Cajal in 1888. In more than one hundred years only a very little percentage of the brain functioning has been discovered, and still much more needs to be explored. Isolated techniques aim at unraveling the system that supports our cognition, nevertheless in order to provide solid evidence in such a field multimodal techniques have arisen, with them we will be able to improve current knowledge about human cognition. Here we focus on the multimodal integration of magnetoencephalography (MEG) and diffusion weighted magnetic resonance imaging. These techniques are sensitive to the magnetic fields emitted by the neuronal currents and to the white matter microstructure, respectively. The combination of such techniques could bring up evidences about structural-functional synergies in the brain information processing and which part of this synergy fails in specific neurological pathologies. In particular, we are interested in the relationship between functional and structural connectivity, and how two integrate this information. We quantify the functional connectivity by studying the phase synchronization or the amplitude correlation between time series obtained by MEG, and so we get an index indicating similarity between neuronal entities, i.e. brain regions. In addition we quantify structural connectivity by performing diffusion tensor estimation from the diffusion weighted images, thus obtaining an indicator of the integrity of the white matter or, if preferred, the strength of the structural connections between regions. These quantifications are then combined following three different approaches, from the lowest to the highest level of integration, in chapters 3, 4 and 5. We finally apply the fused information to the characterization or prediction of mild cognitive impairment, a clinical entity which is considered as an early step in the continuum pathological process of dementia. The dissertation is divided in six chapters. In chapter 1 I introduce connectomics within the fields of neuroimaging and neuroscience. Later in this chapter we describe the objectives of this thesis, and the specific objectives of each of the scientific publications that were produced as result of this work. In chapter 2 I describe the methods for each of the techniques that were employed, namely structural connectivity, resting state functional connectivity, complex brain networks and graph theory, and finally, I describe the clinical condition of mild cognitive impairment and the current state of the art in the search for early biomarkers. In chapters 3, 4 and 5 I have included the scientific publications that were generated along this work. They have been included in in their original format and they contain introduction, materials and methods, results and discussion. All methods that were employed in these papers have been described in chapter 2. Finally, in chapter 6 I summarize all the results from this thesis, both locally for each of the scientific publications and globally for the whole work.

Aplicación de técnicas de inteligencia artificial para contribuir en la detección de microcalcificaciones en mamografía digitalizada

A nivel mundial, el cáncer de mama es el tipo de cáncer más frecuente además de una de las principales causas de muerte entre la población femenina. Actualmente, el método más eficaz para detectar lesiones mamarias en una etapa temprana es la mamografía. Ésta contribuye decisivamente al diagnóstico precoz de esta enfermedad que, si se detecta a tiempo, tiene una probabilidad de curación muy alta. Uno de los principales y más frecuentes hallazgos en una mamografía, son las microcalcificaciones, las cuales son consideradas como un indicador importante de cáncer de mama. En el momento de analizar las mamografías, factores como la capacidad de visualización, la fatiga o la experiencia profesional del especialista radiólogo hacen que el riesgo de omitir ciertas lesiones presentes se vea incrementado. Para disminuir dicho riesgo es importante contar con diferentes alternativas como por ejemplo, una segunda opinión por otro especialista o un doble análisis por el mismo. En la primera opción se eleva el coste y en ambas se prolonga el tiempo del diagnóstico. Esto supone una gran motivación para el desarrollo de sistemas de apoyo o asistencia en la toma de decisiones. En este trabajo de tesis se propone, se desarrolla y se justifica un sistema capaz de detectar microcalcificaciones en regiones de interés extraídas de mamografías digitalizadas, para contribuir a la detección temprana del cáncer demama. Dicho sistema estará basado en técnicas de procesamiento de imagen digital, de reconocimiento de patrones y de inteligencia artificial. Para su desarrollo, se tienen en cuenta las siguientes consideraciones: 1. Con el objetivo de entrenar y probar el sistema propuesto, se creará una base de datos de imágenes, las cuales pertenecen a regiones de interés extraídas de mamografías digitalizadas. 2. Se propone la aplicación de la transformada Top-Hat, una técnica de procesamiento digital de imagen basada en operaciones de morfología matemática. La finalidad de aplicar esta técnica es la de mejorar el contraste entre las microcalcificaciones y el tejido presente en la imagen. 3. Se propone un algoritmo novel llamado sub-segmentación, el cual está basado en técnicas de reconocimiento de patrones aplicando un algoritmo de agrupamiento no supervisado, el PFCM (Possibilistic Fuzzy c-Means). El objetivo es encontrar las regiones correspondientes a las microcalcificaciones y diferenciarlas del tejido sano. Además, con la finalidad de mostrar las ventajas y desventajas del algoritmo propuesto, éste es comparado con dos algoritmos del mismo tipo: el k-means y el FCM (Fuzzy c-Means). Por otro lado, es importante destacar que en este trabajo por primera vez la sub-segmentación es utilizada para detectar regiones pertenecientes a microcalcificaciones en imágenes de mamografía. 4. Finalmente, se propone el uso de un clasificador basado en una red neuronal artificial, específicamente un MLP (Multi-layer Perceptron). El propósito del clasificador es discriminar de manera binaria los patrones creados a partir de la intensidad de niveles de gris de la imagen original. Dicha clasificación distingue entre microcalcificación y tejido sano. ABSTRACT Breast cancer is one of the leading causes of women mortality in the world and its early detection continues being a key piece to improve the prognosis and survival. Currently, the most reliable and practical method for early detection of breast cancer is mammography.The presence of microcalcifications has been considered as a very important indicator ofmalignant types of breast cancer and its detection and classification are important to prevent and treat the disease. However, the detection and classification of microcalcifications continue being a hard work due to that, in mammograms there is a poor contrast between microcalcifications and the tissue around them. Factors such as visualization, tiredness or insufficient experience of the specialist increase the risk of omit some present lesions. To reduce this risk, is important to have alternatives such as a second opinion or a double analysis for the same specialist. In the first option, the cost increases and diagnosis time also increases for both of them. This is the reason why there is a great motivation for development of help systems or assistance in the decision making process. This work presents, develops and justifies a system for the detection of microcalcifications in regions of interest extracted fromdigitizedmammographies to contribute to the early detection of breast cancer. This systemis based on image processing techniques, pattern recognition and artificial intelligence. For system development the following features are considered: With the aim of training and testing the system, an images database is created, belonging to a region of interest extracted from digitized mammograms. The application of the top-hat transformis proposed. This image processing technique is based on mathematical morphology operations. The aim of this technique is to improve the contrast betweenmicrocalcifications and tissue present in the image. A novel algorithm called sub-segmentation is proposed. The sub-segmentation is based on pattern recognition techniques applying a non-supervised clustering algorithm known as Possibilistic Fuzzy c-Means (PFCM). The aim is to find regions corresponding to the microcalcifications and distinguish them from the healthy tissue. Furthermore,with the aim of showing themain advantages and disadvantages this is compared with two algorithms of same type: the k-means and the fuzzy c-means (FCM). On the other hand, it is important to highlight in this work for the first time the sub-segmentation is used for microcalcifications detection. Finally, a classifier based on an artificial neural network such as Multi-layer Perceptron is used. The purpose of this classifier is to discriminate froma binary perspective the patterns built from gray level intensity of the original image. This classification distinguishes between microcalcifications and healthy tissue.

Morphological Functions with Parallel Sets for the Pore Space of X-ray CT Images of Soil Columns

During the last few decades, new imaging techniques like X-ray computed tomography have made available rich and detailed information of the spatial arrangement of soil constituents, usually referred to as soil structure. Mathematical morphology provides a plethora of mathematical techniques to analyze and parameterize the geometry of soil structure. They provide a guide to design the process from image analysis to the generation of synthetic models of soil structure in order to investigate key features of flow and transport phenomena in soil. In this work, we explore the ability of morphological functions built over Minkowski functionals with parallel sets of the pore space to characterize and quantify pore space geometry of columns of intact soil. These morphological functions seem to discriminate the effects on soil pore space geometry of contrasting management practices in a Mediterranean vineyard, and they provide the first step toward identifying the statistical significance of the observed differences.

Linear color correction for multiple illumination changes and non-overlapping cameras

Many image processing methods, such as techniques for people re-identification, assume photometric constancy between different images. This study addresses the correction of photometric variations based upon changes in background areas to correct foreground areas. The authors assume a multiple light source model where all light sources can have different colours and will change over time. In training mode, the authors learn per-location relations between foreground and background colour intensities. In correction mode, the authors apply a double linear correction model based on learned relations. This double linear correction includes a dynamic local illumination correction mapping as well as an inter-camera mapping. The authors evaluate their illumination correction by computing the similarity between two images based on the earth mover's distance. The authors compare the results to a representative auto-exposure algorithm found in the recent literature plus a colour correction one based on the inverse-intensity chromaticity. Especially in complex scenarios the authors’ method outperforms these state-of-the-art algorithms.

New developments in lycopene analysis by spectroscopic and chromatographic techniques, accompanied by mathematical modelling

Comunicación Oral sobre los resultados obtenidos en el estudio de las propiedades del lycopeno presente en el tomate como compuesto bioactivo. Se realizó la identificación y cuantificación por diferentes metodologías experimentales. Se muestran los resultados analíticos comapartivamente con distintas técnicas.

A Dynamically Adaptable Image Processing Application Trading Off Between High Performance, Consumption and Dependability in Real Time

As embedded systems evolve, problems inherent to technology become important limitations. In less than ten years, chips will exceed the maximum allowed power consumption affecting performance, since, even though the resources available per chip are increasing, frequency of operation has stalled. Besides, as the level of integration is increased, it is difficult to keep defect density under control, so new fault tolerant techniques are required. In this demo work, a new dynamically adaptable virtual architecture (ARTICo3) to allow dynamic and context-aware use of resources is implemented in a high performance Wireless Sensor node (HiReCookie) to perform an image processing application.

Study of the effects of moving targets in SAR images for their detection.

Synthetic Aperture Radar’s (SAR) are systems designed in the early 50’s that are capable of obtaining images of the ground using electromagnetic signals. Thus, its activity is not interrupted by adverse meteorological conditions or during the night, as it occurs in optical systems. The name of the system comes from the creation of a synthetic aperture, larger than the real one, by moving the platform that carries the radar (typically a plane or a satellite). It provides the same resolution as a static radar equipped with a larger antenna. As it moves, the radar keeps emitting pulses every 1/PRF seconds —the PRF is the pulse repetition frequency—, whose echoes are stored and processed to obtain the image of the ground. To carry out this process, the algorithm needs to make the assumption that the targets in the illuminated scene are not moving. If that is the case, the algorithm is able to extract a focused image from the signal. However, if the targets are moving, they get unfocused and/or shifted from their position in the final image. There are applications in which it is especially useful to have information about moving targets (military, rescue tasks,studyoftheflowsofwater,surveillanceofmaritimeroutes...).Thisfeatureiscalled Ground Moving Target Indicator (GMTI). That is why the study and the development of techniques capable of detecting these targets and placing them correctly in the scene is convenient. In this document, some of the principal GMTI algorithms used in SAR systems are detailed. A simulator has been created to test the features of each implemented algorithm on a general situation with moving targets. Finally Monte Carlo tests have been performed, allowing us to extract conclusions and statistics of each algorithm.

A study of feature combination for vehicle detection based on image processing

Video analytics play a critical role in most recent traffic monitoring and driver assistance systems. In this context, the correct detection and classification of surrounding vehicles through image analysis has been the focus of extensive research in the last years. Most of the pieces of work reported for image-based vehicle verification make use of supervised classification approaches and resort to techniques, such as histograms of oriented gradients (HOG), principal component analysis (PCA), and Gabor filters, among others. Unfortunately, existing approaches are lacking in two respects: first, comparison between methods using a common body of work has not been addressed; second, no study of the combination potentiality of popular features for vehicle classification has been reported. In this study the performance of the different techniques is first reviewed and compared using a common public database. Then, the combination capabilities of these techniques are explored and a methodology is presented for the fusion of classifiers built upon them, taking into account also the vehicle pose. The study unveils the limitations of single-feature based classification and makes clear that fusion of classifiers is highly beneficial for vehicle verification.

Tag detection for preventing unauthorized face image processing

A new technology is being proposed as a solution to the problem of unintentional facial detection and recognition in pictures in which the individuals appearing want to express their privacy preferences, through the use of different tags. The existing methods for face de-identification were mostly ad hoc solutions that only provided an absolute binary solution in a privacy context such as pixelation, or a bar mask. As the number and users of social networks are increasing, our preferences regarding our privacy may become more complex, leaving these absolute binary solutions as something obsolete. The proposed technology overcomes this problem by embedding information in a tag which will be placed close to the face without being disruptive. Through a decoding method the tag will provide the preferences that will be applied to the images in further stages.

Learning 3D structure from 2D images using LBP features

An automatic machine learning strategy for computing the 3D structure of monocular images from a single image query using Local Binary Patterns is presented. The 3D structure is inferred through a training set composed by a repository of color and depth images, assuming that images with similar structure present similar depth maps. Local Binary Patterns are used to characterize the structure of the color images. The depth maps of those color images with a similar structure to the query image are adaptively combined and filtered to estimate the final depth map. Using public databases, promising results have been obtained outperforming other state-of-the-art algorithms and with a computational cost similar to the most efficient 2D-to-3D algorithms.

New strategies in laser processing of TCOs for light management improvement in thin-film silicon solar cells

Light confinement strategies play a crucial role in the performance of thin-film (TF) silicon solar cells. One way to reduce the optical losses is the texturing of the transparent conductive oxide (TCO) that acts as the front contact. Other losses arise from the mismatch between the incident light spectrum and the spectral properties of the absorbent material that imply that low energy photons (below the bandgap value) are not absorbed, and therefore can not generate photocurrent. Up-conversion techniques, in which two sub-bandgap photons are combined to give one photon with a better matching with the bandgap, were proposed to overcome this problem. In particular, this work studies two strategies to improve light management in thin film silicon solar cells using laser technology. The first one addresses the problem of TCO surface texturing using fully commercial fast and ultrafast solid state laser sources. Aluminum doped Zinc Oxide (AZO) samples were laser processed and the results were optically evaluated by measuring the haze factor of the treated samples. As a second strategy, laser annealing experiments of TCOs doped with rare earth ions are presented as a potential process to produce layers with up-conversion properties, opening the possibility of its potential use in high efficiency solar cells.

A Ground-Truth Video Dataset for the Development and Evaluation of Vision-based Sense-and-Avoid systems

The importance of vision-based systems for Sense-and-Avoid is increasing nowadays as remotely piloted and autonomous UAVs become part of the non-segregated airspace. The development and evaluation of these systems demand flight scenario images which are expensive and risky to obtain. Currently Augmented Reality techniques allow the compositing of real flight scenario images with 3D aircraft models to produce useful realistic images for system development and benchmarking purposes at a much lower cost and risk. With the techniques presented in this paper, 3D aircraft models are positioned firstly in a simulated 3D scene with controlled illumination and rendering parameters. Realistic simulated images are then obtained using an image processing algorithm which fuses the images obtained from the 3D scene with images from real UAV flights taking into account on board camera vibrations. Since the intruder and camera poses are user-defined, ground truth data is available. These ground truth annotations allow to develop and quantitatively evaluate aircraft detection and tracking algorithms. This paper presents the software developed to create a public dataset of 24 videos together with their annotations and some tracking application results.

Material requirements for the adoption of unconventional silicon crystal and wafer growth techniques for high-efficiency solar cells

Silicon wafers comprise approximately 40% of crystalline silicon module cost, and represent an area of great technological innovation potential. Paradoxically, unconventional wafer-growth techniques have thus far failed to displace multicrystalline and Czochralski silicon, despite four decades of innovation. One of the shortcomings of most unconventional materials has been a persistent carrier lifetime deficit in comparison to established wafer technologies, which limits the device efficiency potential. In this perspective article, we review a defect-management framework that has proven successful in enabling millisecond lifetimes in kerfless and cast materials. Control of dislocations and slowly diffusing metal point defects during growth, coupled to effective control of fast-diffusing species during cell processing, is critical to enable high cell efficiencies. To accelerate the pace of novel wafer development, we discuss approaches to rapidly evaluate the device efficiency potential of unconventional wafers from injection-dependent lifetime measurements.

Fault management techniques for systems with SRAM-based FPGAs

Las Field-Programmable Gate Arrays (FPGAs) SRAM se construyen sobre una memoria de configuración de tecnología RAM Estática (SRAM). Presentan múltiples características que las hacen muy interesantes para diseñar sistemas empotrados complejos. En primer lugar presentan un coste no-recurrente de ingeniería (NRE) bajo, ya que los elementos lógicos y de enrutado están pre-implementados (el diseño de usuario define su conexionado). También, a diferencia de otras tecnologías de FPGA, pueden ser reconfiguradas (incluso en campo) un número ilimitado de veces. Es más, las FPGAs SRAM de Xilinx soportan Reconfiguración Parcial Dinámica (DPR), la cual permite reconfigurar la FPGA sin interrumpir la aplicación. Finalmente, presentan una alta densidad de lógica, una alta capacidad de procesamiento y un rico juego de macro-bloques. Sin embargo, un inconveniente de esta tecnología es su susceptibilidad a la radiación ionizante, la cual aumenta con el grado de integración (geometrías más pequeñas, menores tensiones y mayores frecuencias). Esta es una precupación de primer nivel para aplicaciones en entornos altamente radiativos y con requisitos de alta confiabilidad. Este fenómeno conlleva una degradación a largo plazo y también puede inducir fallos instantáneos, los cuales pueden ser reversibles o producir daños irreversibles. En las FPGAs SRAM, los fallos inducidos por radiación pueden aparecer en en dos capas de arquitectura diferentes, que están físicamente superpuestas en el dado de silicio. La Capa de Aplicación (o A-Layer) contiene el hardware definido por el usuario, y la Capa de Configuración contiene la memoria de configuración y la circuitería de soporte. Los fallos en cualquiera de estas capas pueden hacer fracasar el sistema, lo cual puede ser ás o menos tolerable dependiendo de los requisitos de confiabilidad del sistema. En el caso general, estos fallos deben gestionados de alguna manera. Esta tesis trata sobre la gestión de fallos en FPGAs SRAM a nivel de sistema, en el contexto de sistemas empotrados autónomos y confiables operando en un entorno radiativo. La tesis se centra principalmente en aplicaciones espaciales, pero los mismos principios pueden aplicarse a aplicaciones terrenas. Las principales diferencias entre ambas son el nivel de radiación y la posibilidad de mantenimiento. Las diferentes técnicas para la gestión de fallos en A-Layer y C-Layer son clasificados, y sus implicaciones en la confiabilidad del sistema son analizados. Se proponen varias arquitecturas tanto para Gestores de Fallos de una capa como de doble-capa. Para estos últimos se propone una arquitectura novedosa, flexible y versátil. Gestiona las dos capas concurrentemente de manera coordinada, y permite equilibrar el nivel de redundancia y la confiabilidad. Con el objeto de validar técnicas de gestión de fallos dinámicas, se desarrollan dos diferentes soluciones. La primera es un entorno de simulación para Gestores de Fallos de C-Layer, basado en SystemC como lenguaje de modelado y como simulador basado en eventos. Este entorno y su metodología asociada permite explorar el espacio de diseño del Gestor de Fallos, desacoplando su diseño del desarrollo de la FPGA objetivo. El entorno incluye modelos tanto para la C-Layer de la FPGA como para el Gestor de Fallos, los cuales pueden interactuar a diferentes niveles de abstracción (a nivel de configuration frames y a nivel físico JTAG o SelectMAP). El entorno es configurable, escalable y versátil, e incluye capacidades de inyección de fallos. Los resultados de simulación para algunos escenarios son presentados y comentados. La segunda es una plataforma de validación para Gestores de Fallos de FPGAs Xilinx Virtex. La plataforma hardware aloja tres Módulos de FPGA Xilinx Virtex-4 FX12 y dos Módulos de Unidad de Microcontrolador (MCUs) de 32-bits de propósito general. Los Módulos MCU permiten prototipar Gestores de Fallos de C-Layer y A-Layer basados en software. Cada Módulo FPGA implementa un enlace de A-Layer Ethernet (a través de un switch Ethernet) con uno de los Módulos MCU, y un enlace de C-Layer JTAG con el otro. Además, ambos Módulos MCU intercambian comandos y datos a través de un enlace interno tipo UART. Al igual que para el entorno de simulación, se incluyen capacidades de inyección de fallos. Los resultados de pruebas para algunos escenarios son también presentados y comentados. En resumen, esta tesis cubre el proceso completo desde la descripción de los fallos FPGAs SRAM inducidos por radiación, pasando por la identificación y clasificación de técnicas de gestión de fallos, y por la propuesta de arquitecturas de Gestores de Fallos, para finalmente validarlas por simulación y pruebas. El trabajo futuro está relacionado sobre todo con la implementación de Gestores de Fallos de Sistema endurecidos para radiación. ABSTRACT SRAM-based Field-Programmable Gate Arrays (FPGAs) are built on Static RAM (SRAM) technology configuration memory. They present a number of features that make them very convenient for building complex embedded systems. First of all, they benefit from low Non-Recurrent Engineering (NRE) costs, as the logic and routing elements are pre-implemented (user design defines their connection). Also, as opposed to other FPGA technologies, they can be reconfigured (even in the field) an unlimited number of times. Moreover, Xilinx SRAM-based FPGAs feature Dynamic Partial Reconfiguration (DPR), which allows to partially reconfigure the FPGA without disrupting de application. Finally, they feature a high logic density, high processing capability and a rich set of hard macros. However, one limitation of this technology is its susceptibility to ionizing radiation, which increases with technology scaling (smaller geometries, lower voltages and higher frequencies). This is a first order concern for applications in harsh radiation environments and requiring high dependability. Ionizing radiation leads to long term degradation as well as instantaneous faults, which can in turn be reversible or produce irreversible damage. In SRAM-based FPGAs, radiation-induced faults can appear at two architectural layers, which are physically overlaid on the silicon die. The Application Layer (or A-Layer) contains the user-defined hardware, and the Configuration Layer (or C-Layer) contains the (volatile) configuration memory and its support circuitry. Faults at either layers can imply a system failure, which may be more ore less tolerated depending on the dependability requirements. In the general case, such faults must be managed in some way. This thesis is about managing SRAM-based FPGA faults at system level, in the context of autonomous and dependable embedded systems operating in a radiative environment. The focus is mainly on space applications, but the same principles can be applied to ground applications. The main differences between them are the radiation level and the possibility for maintenance. The different techniques for A-Layer and C-Layer fault management are classified and their implications in system dependability are assessed. Several architectures are proposed, both for single-layer and dual-layer Fault Managers. For the latter, a novel, flexible and versatile architecture is proposed. It manages both layers concurrently in a coordinated way, and allows balancing redundancy level and dependability. For the purpose of validating dynamic fault management techniques, two different solutions are developed. The first one is a simulation framework for C-Layer Fault Managers, based on SystemC as modeling language and event-driven simulator. This framework and its associated methodology allows exploring the Fault Manager design space, decoupling its design from the target FPGA development. The framework includes models for both the FPGA C-Layer and for the Fault Manager, which can interact at different abstraction levels (at configuration frame level and at JTAG or SelectMAP physical level). The framework is configurable, scalable and versatile, and includes fault injection capabilities. Simulation results for some scenarios are presented and discussed. The second one is a validation platform for Xilinx Virtex FPGA Fault Managers. The platform hosts three Xilinx Virtex-4 FX12 FPGA Modules and two general-purpose 32-bit Microcontroller Unit (MCU) Modules. The MCU Modules allow prototyping software-based CLayer and A-Layer Fault Managers. Each FPGA Module implements one A-Layer Ethernet link (through an Ethernet switch) with one of the MCU Modules, and one C-Layer JTAG link with the other. In addition, both MCU Modules exchange commands and data over an internal UART link. Similarly to the simulation framework, fault injection capabilities are implemented. Test results for some scenarios are also presented and discussed. In summary, this thesis covers the whole process from describing the problem of radiationinduced faults in SRAM-based FPGAs, then identifying and classifying fault management techniques, then proposing Fault Manager architectures and finally validating them by simulation and test. The proposed future work is mainly related to the implementation of radiation-hardened System Fault Managers.