Simultaneous localization and mapping is the technique to construct a 3D map of unknown environment. With the increasing popularity of RGB-depth (RGB-D) sensors such as the Microsoft Kinect, there have been much research on capturing and reconstructing 3D environments using a movable RGB-D sensor. The key process behind these kinds of simultaneous location and mapping (SLAM) systems is the iterative closest point or ICP algorithm, which is an iterative algorithm that can estimate the rigid movement of the camera based on the captured 3D point clouds. While ICP is a well-studied algorithm, it is problematic when it is used in scanning large planar regions such as wall surfaces in a room. The lack of depth variations on planar surfaces makes the global alignment an ill-conditioned problem. In this thesis, we present a novel approach for registering 3D point clouds by combining both color and depth information. Instead of directly searching for point correspondences among 3D data, the proposed method first extracts features from the RGB images, and then back-projects the features to the 3D space to identify more reliable correspondences. These color correspondences form the initial input to the ICP procedure which then proceeds to refine the alignment. Experimental results show that our proposed approach can achieve better accuracy than existing SLAMs in reconstructing indoor environments with large planar surfaces.

Avec un encombrement minimal et rendement de 99% les transformateurs planars haute fréquence sont des solutions technologiques importantes pour les systèmes de conversion et de conditionnement de l'énergie électrique en électronique de puissance. L'objectif de l'étude consiste à rechercher un modèle du composant permettant à la fois de simuler son comportement dans un circuit d'électronique de puissance et d'optimiser ses performances lors de sa conception. Les modèles proposés prennent en compte les propriétés physiques et géométriques du composant. Les validations sont basées sur des confrontations avec les mesures. Une autre motivation forte de ce travail est que les modélisations numériques doivent prendre en compte simultanément les effets des courants induits et les capacités parasites, ce qui conduit, en haute fréquence, à des besoins prohibitifs en termes de capacité mémoire ou de temps de calcul. En se limitant aux transformateurs planar. La modélisation (dite LEEC) présentée ici s'appuie sur une discrétisation à échelle intermédiaire : spire par spire, c'est-à-dire couche par couche. Elle assemble deux approches analytiques déjà introduites par l'équipe : la première traite les aspects électrocinétiques (incluant les courants induits) et magnétiques et l'autre les aspects électrostatiques. Les circuits à constantes localisées basés sur la méthode LEEC montrent un très bon accord avec toutes les mesures jusqu'à 40 MHz. Des outils numériques sont aussi développés pour faciliter l'obtention de ces différents circuits en partant : soit de la description du composant, soit de mesures d'impédances. Pour compléter, la représentation des capacités d'un transformateur quelconque, développée en plusieurs étapes par l'équipe, est résumée et des précautions expérimentales originales appliquées à l'impédancemétrie sont exposées dans le détail. D'autres travaux visant à élargir le champ d'application de la méthode LEEC sont également présentés.

This dissertation reports on the design and implementation of a high-frequency, tri-axial capacitive resonant gyroscopes integrated on a single chip. The components that construct tri-axial rotation sensing consist of a yaw, a pitch and a roll device. The yaw-rate gyroscope has a wide bandwidth and a large full-scale range, and operates at a mode-matched condition with DC polarization voltage of 10V without frequency tuning requirement. The large bandwidth of 3kHz and expected full-scale range over 30,000˚/sec make the device exhibit fast rate response for rapid motion sensing application. For the pitch-and-roll rate sensing, an in-plane drive-mode and two orthogonal out-of-plane sense-modes are employed. The rotation-rate sensing from lateral axes is performed by mode-matching the in-plane drive-mode with out-of-plane sense-modes to detect Coriolis-force induced deflection of the resonant mass. To compensate process variations and thickness deviations in the employed silicon-on-insulator (SOI) substrates, large electrostatic frequency tunings of both the drive and sense modes are realized. A revised high aspect ratio combined polysilicon and silicon (HARPSS) process is developed to resolve the Coriolis response that exists toward out-of-plane direction while drive-mode exists on in-plane, and tune individual frequencies with minimal interference to unintended modes. To conclude and overcome the performance limitation, design optimization of high-frequency tri-axial gyroscopes is suggested. Q-factor enhancement through reduction of thermoelastic damping (TED) and optimizations of physical dimensions are suggested for the yaw disk gyroscope. For the pitch-and-roll gyroscope, scaling property of physical dimension and its subsequent performance enhancement are analyzed.

In this thesis, we investigate the formulation, optimization and ambiguities in single-image 3D surface reconstruction from geometric and photometric constraints. We examine linear, discrete and quadratic constraints for shape from planar curves, shape from texture, and shape from shading. The problem of recovering 3D shape from the projection of planar curves on a surface is strongly motivated by perception studies. Applications include single-view modeling and uncalibrated structured light. When the curves intersect, the problem leads to a linear system for which a direct least-squares method is sensitive to noise. We derive a more stable solution and show examples where the same method produces plausible surfaces from the projection of parallel (non-intersecting) planar cross sections. The problem of reconstructing a smooth surface under constraints that have discrete ambiguities arise in areas such as shape from texture, shape from shading, photometric stereo and shape from defocus. While the problem is computationally hard, heuristics based on semidefinite programming may reveal the shape of the surface. Finally, we examine the shape from shading problem without boundary conditions as a polynomial system. This formulation allows, in generic cases, a complete solution for ideal polyhedral objects. For the general case we propose a semidefinite programming relaxation procedure, and an exact line search iterative procedure with a new smoothness term that favors folds at edges. We use this numerical technique to inspect shading ambiguities.

A novel photonic frequency discriminator has been developed. The discriminator utilizes a Mach Zehnder interferometer-assisted ring resonator to achieve enhanced linearity. A numerical frequency-domain two-tone test is performed to evaluate the unique design of the discriminator, particularly for suppression of the third order intermodulation distortion. The discriminator is switchable between linear-intensity and linear-field regimes by adjusting a phase delay on one arm of the Mach Zehnder interferometer. Through the simulation, the linear<intensity discriminator is shown to be advantageous. The discriminator is an optical ring resonator-Mach Zehnder interferometer synthesized passive filter. The ring resonator is made of Arsenic trisulfide (As2S3) and the bus waveguide is a Titanium<diffused Lithium niobate (LiNbO3) waveguide. This As2S3 ring-on-Ti:LiNbO3 hybrid structure offers electro-optic tunability of the device owing to a strong electro-optic effect of the substrate material. A large optical confinement factor achieved by vertical integration of the As2S3 strip waveguide on a LiNbO3 substrate enables a low loss ring resonator. The Mach Zehnder interferometer is formed by the optical path length difference of the birefringent LiNbO3 substrate instead of a physical Y-branch structure, which makes the fabrication tolerances relaxed. In order for this highly birefringent device to be characterized, each polarization mode must be measured separately. A novel algorithm which can measure the wavelength-swept Jones matrix including its phase response is devised. The efficacy of the algorithm is demonstrated by characterizing a ring resonator. Finally, the fabricated discriminator is fully characterized using the algorithm.

Many graphs arising in practice can be represented in a concise and intuitive way that conveys their structure. For example: A planar graph can be represented in the plane with points for vertices and non-crossing curves for edges. An interval graph can be represented on the real line with intervals for vertices and intersection of intervals representing edges. The concept of ``simultaneity'' applies for several types of graphs: the idea is to find representations for two graphs that share some common vertices and edges, and ensure that the common vertices and edges are represented the same way. Simultaneous representation problems arise in any situation where two related graphs should be represented consistently. A main instance is for temporal relationships, where an old graph and a new graph share some common parts. Pairs of related graphs arise in many other situations. For example, two social networks that share some members; two schedules that share some events, overlap graphs of DNA fragments of two similar organisms, circuit graphs of two adjacent layers on a computer chip etc. In this thesis, we study the simultaneous representation problem for several graph classes. For planar graphs the problem is defined as follows. Let G1 and G2 be two graphs sharing some vertices and edges. The simultaneous planar embedding problem asks whether there exist planar embeddings (or drawings) for G1 and G2 such that every vertex shared by the two graphs is mapped to the same point and every shared edge is mapped to the same curve in both embeddings. Over the last few years there has been a lot of work on simultaneous planar embeddings, which have been called `simultaneous embeddings with fixed edges'. A major open question is whether simultaneous planarity for two graphs can be tested in polynomial time. We give a linear-time algorithm for testing the simultaneous planarity of any two graphs that share a 2-connected subgraph. Our algorithm also extends to the case of k planar graphs, where each vertex [edge] is either common to all graphs or belongs to exactly one of them. Next we introduce a new notion of simultaneity for intersection graph classes (interval graphs, chordal graphs etc.) and for comparability graphs. For interval graphs, the problem is defined as follows. Let G1 and G2 be two interval graphs sharing some vertices I and the edges induced by I. G1 and G2 are said to be `simultaneous interval graphs' if there exist interval representations of G1 and G2 such that any vertex of I is assigned to the same interval in both the representations. The `simultaneous representation problem' for interval graphs asks whether G1 and G2 are simultaneous interval graphs. The problem is defined in a similar way for other intersection graph classes. For comparability graphs and any intersection graph class, we show that the simultaneous representation problem for the graph class is equivalent to a graph augmentation problem: given graphs G1 and G2, sharing vertices I and the corresponding induced edges, do there exist edges E' between G1-I and G2-I such that the graph G1 U G_2 U E' belongs to the graph class. This equivalence implies that the simultaneous representation problem is closely related to other well-studied classes in the literature, namely, sandwich graphs and probe graphs. We give efficient algorithms for solving the simultaneous representation problem for interval graphs, chordal graphs, comparability graphs and permutation graphs. Further, our algorithms for comparability and permutation graphs solve a more general version of the problem when there are multiple graphs, any two of which share the same common graph. This version of the problem also generalizes probe graphs.

Les composants magnétiques en technologie planar répondent aux exigences actuelles de l'Electronique de Puissance (EP), à savoir la montée en fréquence de commutation des structures d'EP et la réduction du volume des convertisseurs. La première tendance impose des contraintes fortes en termes de compatibilité électromagnétique (CEM) des équipements. Ces dernières doivent être prises en compte par les ingénieurs dès la phase conception des convertisseurs en se basant sur des modèles fiables, peu développés pour les composants planar dans la littérature scientifique. Ce travail de thèse porte ainsi sur la modélisation des composants planar pour applications aux filtres CEM. Différentes méthodes sont développées au cours de cette thèse pour arriver à évaluer de manière fine les éléments parasites des inductances planar de mode commun : capacités parasites et inductances de fuite. Une partie du travail a porté sur la modélisation par circuits équivalents du comportement fréquentiel des inductances de MC. Une approche automatisée, basée sur un algorithme de fitting a ainsi été développée pour élaborer des circuits équivalents fiables et robustes. Des approches analytiques (Décomposition du Champ Electrique) et semi-analytiques (Fonctions de Green) ont aussi été proposées pour évaluer les valeurs des éléments parasites. La dernière partie de la thèse est plus orientée conception, avec la réalisation de deux structures de composants innovantes, la première se basant sur une technique de compensation des capacités parasites à l'aide d'éléments parasites structuraux et la seconde sur l'association de deux noyaux magnétiques, possédant matériaux et géométries différentes

Planar electromagnetic sensing system has been used as one of the NDT methods to evaluate the material properties i.e., to evaluate near-surface properties such as conductivity, permeability and dielectric properties. The applications of planar electromagnetic sensors will depend on both the characteristic of the sensor type chosen and also the characteristic of material under test. Conventional planar interdigital sensors and novel planar interdigital sensors have been designed, fabricated and tested for detection of dangerous marine biotoxins in seafood. Our main objective is to sense the presence of dangerous contaminated acid in mussels and other seafoods. Initial studies were conducted with three peptide derivatives namely Sarcosine, Proline and Hydroxylproline. These three chemicals are structurally closely related to our target molecule (domoic acid). The initial results have shown that all sensors respond very well to the chemicals and it is possible to discriminate the different chemicals from the output of the sensor. Novel interdigital sensors have shown better sensitivity measurement compared to conventional interdigital sensors. The novel interdigital sensors were then being tested with three seafood products. Results from the analysis have shown that novel interdigital sensor with configuration #1 (Sensor_1) has better sensitivity compared to other sensors. Sensor_1 has been chosen for experiment using proline and mussels. The changes in sensor sensitivity were analysed with mussels before and after adding the proline. The presence of proline on the mussel surface and also injected proline to the mussel samples were clearly detected by Sensor_1. Further experiment was conducted with small amount of domoic acid (0.5 µg to 5.0 µg) injected to a mussel and it was found that Sensor_1 was able to detect small amount of domoic acid (1.0 µg) injected into the mussel sample. Sensor_1 was able to detect approximately 12.6 µg/g of domoic acid in mussel meat. Three threshold levels of particular sample thickness have been established for detection of domoic acid. The first prototype of a low cost sensing system known as SIT (Seafood Inspection Tool) has been developed. The outcomes from the experiments provide chances of opportunity for further research in developing a low cost miniature type of sensors for reliable sensing system for commercial use.

Planar electromagnetic sensing system has been used as one of the NDT methods to evaluate the material properties i.e., to evaluate near-surface properties such as conductivity, permeability and dielectric properties. The applications of planar electromagnetic sensors will depend on both the characteristic of the sensor type chosen and also the characteristic of material under test. Conventional planar interdigital sensors and novel planar interdigital sensors have been designed, fabricated and tested for detection of dangerous marine biotoxins in seafood. Our main objective is to sense the presence of dangerous contaminated acid in mussels and other seafoods. Initial studies were conducted with three peptide derivatives namely Sarcosine, Proline and Hydroxylproline. These three chemicals are structurally closely related to our target molecule (domoic acid). The initial results have shown that all sensors respond very well to the chemicals and it is possible to discriminate the different chemicals from the output of the sensor. Novel interdigital sensors have shown better sensitivity measurement compared to conventional interdigital sensors. The novel interdigital sensors were then being tested with three seafood products. Results from the analysis have shown that novel interdigital sensor with configuration #1 (Sensor_1) has better sensitivity compared to other sensors. Sensor_1 has been chosen for experiment using proline and mussels. The changes in sensor sensitivity were analysed with mussels before and after adding the proline. The presence of proline on the mussel surface and also injected proline to the mussel samples were clearly detected by Sensor_1. Further experiment was conducted with small amount of domoic acid (0.5 µg to 5.0 µg) injected to a mussel and it was found that Sensor_1 was able to detect small amount of domoic acid (1.0 µg) injected into the mussel sample. Sensor_1 was able to detect approximately 12.6 µg/g of domoic acid in mussel meat. Three threshold levels of particular sample thickness have been established for detection of domoic acid. The first prototype of a low cost sensing system known as SIT (Seafood Inspection Tool) has been developed. The outcomes from the experiments provide chances of opportunity for further research in developing a low cost miniature type of sensors for reliable sensing system for commercial use.

Planar electromagnetic sensing system has been used as one of the NDT methods to evaluate the material properties i.e., to evaluate near-surface properties such as conductivity, permeability and dielectric properties. The applications of planar electromagnetic sensors will depend on both the characteristic of the sensor type chosen and also the characteristic of material under test. Conventional planar interdigital sensors and novel planar interdigital sensors have been designed, fabricated and tested for detection of dangerous marine biotoxins in seafood. Our main objective is to sense the presence of dangerous contaminated acid in mussels and other seafoods. Initial studies were conducted with three peptide derivatives namely Sarcosine, Proline and Hydroxylproline. These three chemicals are structurally closely related to our target molecule (domoic acid). The initial results have shown that all sensors respond very well to the chemicals and it is possible to discriminate the different chemicals from the output of the sensor. Novel interdigital sensors have shown better sensitivity measurement compared to conventional interdigital sensors. The novel interdigital sensors were then being tested with three seafood products. Results from the analysis have shown that novel interdigital sensor with configuration #1 (Sensor_1) has better sensitivity compared to other sensors. Sensor_1 has been chosen for experiment using proline and mussels. The changes in sensor sensitivity were analysed with mussels before and after adding the proline. The presence of proline on the mussel surface and also injected proline to the mussel samples were clearly detected by Sensor_1. Further experiment was conducted with small amount of domoic acid (0.5 µg to 5.0 µg) injected to a mussel and it was found that Sensor_1 was able to detect small amount of domoic acid (1.0 µg) injected into the mussel sample. Sensor_1 was able to detect approximately 12.6 µg/g of domoic acid in mussel meat. Three threshold levels of particular sample thickness have been established for detection of domoic acid. The first prototype of a low cost sensing system known as SIT (Seafood Inspection Tool) has been developed. The outcomes from the experiments provide chances of opportunity for further research in developing a low cost miniature type of sensors for reliable sensing system for commercial use.