The primary goal of this dissertation is to develop point-based rigid and non-rigid image registration methods that have better accuracy than existing methods. We first present point-based PoIRe, which provides the framework for point-based global rigid registrations. It allows a choice of different search strategies including (a) branch-and-bound, (b) probabilistic hill-climbing, and (c) a novel hybrid method that takes advantage of the best characteristics of the other two methods. We use a robust similarity measure that is insensitive to noise, which is often introduced during feature extraction. We show the robustness of PoIRe using it to register images obtained with an electronic portal imaging device (EPID), which have large amounts of scatter and low contrast. To evaluate PoIRe we used (a) simulated images and (b) images with fiducial markers; PoIRe was extensively tested with 2D EPID images and images generated by 3D Computer Tomography (CT) and Magnetic Resonance (MR) images. PoIRe was also evaluated using benchmark data sets from the blind retrospective evaluation project (RIRE). We show that PoIRe is better than existing methods such as Iterative Closest Point (ICP) and methods based on mutual information. We also present a novel point-based local non-rigid shape registration algorithm. We extend the robust similarity measure used in PoIRe to non-rigid registrations adapting it to a free form deformation (FFD) model and making it robust to local minima, which is a drawback common to existing non-rigid point-based methods. For non-rigid registrations we show that it performs better than existing methods and that is less sensitive to starting conditions. We test our non-rigid registration method using available benchmark data sets for shape registration. Finally, we also explore the extraction of features invariant to changes in perspective and illumination, and explore how they can help improve the accuracy of multi-modal registration. For multimodal registration of EPID-DRR images we present a method based on a local descriptor defined by a vector of complex responses to a circular Gabor filter.

The main result of this thesis is a deep learning model named BearNet, which can be trained to detect an arbitrary amount of objects as a set of points. The model is trained using the Weighted Hausdorff distance as loss function. BearNet has been applied and tested on two problems from the industry. These are: From an intensity image, detect two pocket points of an EU-pallet which an autonomous forklift could utilize when determining where to insert its forks. From a depth image, detect the start, bend and end points of a straw attached to a juice package, in order to help determine if the straw has been attached correctly. In the development process of BearNet I took inspiration from the designs of U-Net, UNet++ and a high resolution network named HRNet. Further, I used a dataset containing RGB-images from a surveillance camera located inside a mall, on which the aim was to detect head positions of all pedestrians. In an attempt to reproduce a result from another study, I found that the mall dataset suffers from training set contamination when a model is trained, validated, and tested on it with random sampling. Hence, I propose that the mall dataset is evaluated with a sequential data split strategy, to limit the problem. I found that the BearNet architecture is well suited for both the EU-pallet and straw datasets, and that it can be successfully used on either RGB,  intensity or depth images. On the EU-pallet and straw datasets, BearNet consistently produces point estimates within five and six pixels of ground truth, respectively. I also show that the straw dataset only constitutes a small subset of all the challenges that exist in the problem domain related to the attachment of a straw to a juice package, and that one therefore cannot train a robust deep learning model on it. As an example of this, models trained on the straw dataset cannot correctly handle samples in which there is no straw visible.

This study examines the efficacy of Sentinel-1 Radial Velocity (RVL) imagery at determining the position of ocean current front features, using the Gulf Stream (GS) and Agulhas Current (AC) as case studies. Fronts derived from RVL imagery are compared to fronts derived from Sea Surface Temperature (SST) imagery, specifically Multi-scale Ultra-high Resolution Sea Surface Temperature Analysis (MURSST) data. In the case of the GS, front locations from the Naval Oceanographic Office (NAVOCEANO) were also used for comparison. Only the northern walls of ocean current features are considered in this study, which is broken into three main steps: Preprocessing, front extraction, and front comparison. First, RVL imagery is selected from Sentinel-1 ocean products, preprocessed to remove antenna mispointing artifacts, and all products from the same orbit are combined into a single swath. Second, front features are extracted from both the RVL and MURSST imagery using a ridge detection algorithm, the main ocean current is chosen from all ridge features using a ranking algorithm, and the northern wall of this current is extracted. Third, the RVL, SST, and in the case of the GS, NAVOCEANO GS locations, features are compared using a symmetric Hausdorff Distance (HD) measure, and Mean Hausdorff Distance (MHD). In some cases, the automatic front extraction failed by either misclassifying an eddy or similar ocean feature as the ocean current in either the RVL or SST image or failed to extract the entire length of the front visible within the image. All the SST and RVL fronts were classified manually to determine the success rate of the automatic front extraction and to exclude failed front extractions from the analysis, as they are not accurate representations of the SST and RVL data’s ability to detect fronts. In special cases, the RVL image itself does not detect the entire ocean current, such that there are noticeable gaps in the ocean current. Similarly, in special cases the MURSST does not detect the entire ocean current. The automatic front extraction succeeded 65% of the time, including the special cases. The results demonstrated that RVL products were effective at determining the location of ocean fronts where the angle of the front's normal vector is within approximately 40° of the sensor’s azimuthal heading. A mean HD of 31.9 km and a mean MHD of 13.2 km was calculated for all front pairs over all study areas. The RVL fronts appeared consistently to the north of the SST fronts, with an average offset of 25.4 km between the centroids of the SST and RVL fronts. Positive correlations were noted between cloud coverage and MURSST error in both study regions. Several RVL images detected ocean currents in regions of high MURSST error where the MURSST did not detect the ocean currents, suggesting that RVL may provide more accuracy than SST-based products in clouded regions where there is no auxiliary data.

No description available.

L’objet de cette thèse est d’étudier plusieurs modèles probabilistes reliant les marches aléatoires et les arbres aléatoires issus de processus de branchement critiques.Dans la première partie, nous nous intéressons au modèle de marche aléatoire à valeurs dans un réseau euclidien et indexée par un arbre de Galton–Watson critique conditionné par la taille. Sous certaines hypothèses sur la loi de reproduction critique et la loi de saut centrée, nous obtenons, dans toutes les dimensions, la vitesse de croissance asymptotique du nombre de points visités par cette marche, lorsque la taille de l’arbre tend vers l’infini. Ces résultats nous permettent aussi de décrire le comportement asymptotique du nombre de points visités par une marche aléatoire branchante, quand la taille de la population initiale tend vers l’infini. Nous traitons également en parallèle certains cas où la marche aléatoire possède une dérive constante non nulle.Dans la deuxième partie, nous nous concentrons sur les propriétés fractales de la mesure harmonique des grands arbres de Galton–Watson critiques. On comprend par mesure harmonique la distribution de sortie, hors d’une boule centrée à la racine de l’arbre, d’une marche aléatoire simple sur cet arbre. Lorsque la loi de reproduction critique appartient au domaine d’attraction d’une loi stable, nous prouvons que la masse de la mesure harmonique est asymptotiquement concentrée sur une partie de la frontière, cette partie ayant une taille négligeable par rapport à celle de la frontière. En supposant que la loi de reproduction critique a une variance finie, nous arrivons à évaluer la masse de la mesure harmonique portée par un sommet de la frontière choisi uniformément au hasard. / The aim of this Ph. D. thesis is to study several probabilistic models linking the random walks and the random trees arising from critical branching processes.In the first part, we consider the model of random walk taking values in a Euclidean lattice and indexed by a critical Galton–Watson tree conditioned by the total progeny. Under some assumptions on the critical offspring distribution and the centered jump distribution, we obtain, in all dimensions, the asymptotic growth rate of the range of this random walk, when the size of the tree tends to infinity. These results also allow us to describe the asymptotic behavior of the range of a branching random walk, when the size of the initial population goes to infinity. In parallel, we treat likewise some cases where the random walk has a non-zero constant drift.In the second part, we focus on the fractal properties of the harmonic measure on large critical Galton–Watson trees. By harmonic measure, we mean the exit distribution from a ball centered at the root of the tree by simple random walk on this tree. If the critical offspring distribution is in the domain of attraction of a stable distribution, we prove that the mass of the harmonic measure is asymptotically concentrated on a boundary subset of negligible size with respect to that of the boundary. Assuming that the critical offspring distribution has a finite variance, we are able to calculate the mass of the harmonic measure carried by a random vertex uniformly chosen from the boundary.

Nous étudions les propriétés métriques locales des ensembles de niveau des applicationshorizontalement différentiables entre des groupes de Carnot, c'est-à-dire différentiable par rapport à la structure sous-riemannienne intrinsèque.Nous considérons des applications dont la différentielle horizontale est surjective,et notre étude peut être vue comme une généralisation du théorème des fonctions implicites pour les groupes de Carnot.Tout d'abord, nous présentons deux notions de tangence dans les groupes de Carnot:la première basée sur la condition de platitude au sens de Reifenberg et la deuxième issue de l'analyse convexe classique.Nous montrons que dans les deux cas, l'espace tangent à un ensemble de niveau coïncide avec le noyau de la différentielle horizontale.Nous montrons que cette condition de tangence caractérise en fait les ensembles de niveaudits ‘co-abéliens', c'est-à-dire ceux pour lesquels l'espace d'arrivée est abélien, et qu'une telle caractérisation n'est pas vraie en général.Ce résultat sur les espaces tangents a plusieurs conséquences remarquables.La plus importante est que la dimension de Hausdorff des ensembles de niveau est celle à laquelle l'on s'attend.Nous montrons également la connectivité locale des ensembles de niveau, et le fait que les ensembles de niveau de dimension 1 sont topologiquement des arcs simples.Pour les ensembles de niveau de dimension 1 nous trouvons une formule de l'aire qui permet d'exprimer la mesure de Hausdorff en termes d'intégrales de Stieltjes généralisées.Ensuite, nous menons une étude approfondie du cas particulier des ensembles de niveau dans les groupes d'Heisenberg.Nous montrons que les ensembles de niveau sont topologiquement équivalents à leurs espaces tangents.Il s'avère que la mesure de Hausdorff des ensembles de niveau de codimension élevée est souvent irrégulière, étant, par exemple, localement nulle ou infinie.Nous présentons une condition simple de régularité supplémentaire pour une application pour assurer la régularité au sens d'Ahlfors des ses ensembles de niveau.Parmi d'autres résultats, nous obtenons une nouvelle caractérisation généraledes graphes Lipschitziens associés à une décomposition en produit semi-direct d'un groupe de Carnot.Nous traitons, en particulier, le cas des groupes de Carnot dont le nombre de stratesest plus grand que $2$.Cette caractérisation nous permet de déduire une nouvelle caractérisation des ensemblesde niveau co-abéliens qui admettent une représentation en tant que graphe. / Metric properties of level sets of differentiable maps on Carnot groupsAbstract.We investigate the local metric properties of level sets of mappings defined between Carnot groups that are horizontally differentiable, i.e.with respect to the intrinsic sub-Riemannian structure. We focus on level sets of mapping having a surjective differential,thus, our study can be seen as an extension of implicit function theorem for Carnot groups.First, we present two notions of tangency in Carnot groups: one based on Reifenberg's flatness condition and another coming from classical convex analysis.We show that for both notions, the tangents to level sets coincide with the kernels of horizontal differentials.Furthermore, we show that this kind of tangency characterizes the level sets called ``co-abelian'', i.e.for which the target space is abelian andthat such a characterization may fail in general.This tangency result has several remarkable consequences.The most important one is that the Hausdorff dimension of the level sets is the expected one. We also show the local connectivity of level sets and, the fact that level sets of dimension one are topologically simple arcs.Again for dimension one level set, we find an area formula that enables us to compute the Hausdorff measurein terms of generalized Stieltjes integrals.Next, we study deeply a particular case of level sets in Heisenberg groups. We show that the level sets in this case are topologically equivalent to their tangents.It turns out that the Hausdorff measure of high-codimensional level sets behaves wildly, for instance, it may be zero or infinite.We provide a simple sufficient extra regularity condition on mappings that insures Ahlfors regularity of level sets.Among other results, we obtain a new general characterization of Lipschitz graphs associated witha semi-direct splitting of a Carnot group of arbitrary step.We use this characterization to derive a new characterization of co-ablian level sets that can be represented as graphs.

No description available.

Cette thèse est consacrée à l'étude de la régularité des solutions des équations de Monge-Ampère complexes ainsi que des équations hessiennes complexes dans un domaine borné de Cn. Dans le premier chapitre, on donne des rappels sur la théorie du pluripotentiel. Dans le deuxième chapitre, on étudie le module de continuité des solutions du problème de Dirichlet pour les équations de Monge-Ampère lorsque le second membre est une mesure à densité continue par rapport à la mesure de Lebesgue dans un domaine strictement hyperconvexe lipschitzien. Dans le troisième chapitre, on prouve la continuité hölderienne des solutions de ce problème pour certaines mesures générales. Dans le quatrième chapitre, on considère le problème de Dirichlet pour les équations hessiennes complexes plus générales où le second membre dépend de la fonction inconnue. On donne une estimation précise du module de continuité de la solution lorsque la densité est continue. De plus, si la densité est dans Lp , on démontre que la solution est Hölder-continue jusqu'au bord. / In this thesis we study the regularity of solutions to the Dirichlet problem for complex Monge-Ampère equations and also for complex Hessian equations in a bounded domain of Cn. In the first chapter, we give basic facts in pluripotential theory. In the second chapter, we study the modulus of continuity of solutions to the Dirichlet problem for complex Monge-Ampère equations when the right hand side is a measure with continuous density with respect to the Lebesgue measure in a bounded strongly hyperconvex Lipschitz domain. In the third chapter, we prove the Hölder continuity of solutions to this problem for some general measures. In the fourth chapter, we consider the Dirichlet problem for complex Hessian equations when the right hand side depends on the unknown function. We give a sharp estimate of the modulus of continuity of the solution as the density is continuous. Moreover, for the case of Lp-density we demonstrate that the solution is Hölder continuous up to the boundary.

Tato diplomová práce se zabývá teorií fraktálů a popisuje patričné potíže při zavedení pojmu fraktál. Dále se v práci navrhuje několik metod, které se použijí na aproximaci fraktálních dimenzí různých množin zobrazených na zařízeních s konečným rozlišením. Tyto metody se otestují na takových množinách, jejichž dimenze známe, a na závěr se výsledky porovnávají.

Fractals is a relatively new mathematical topic which received thorough treatment only starting with 1960's. Fractals can be observed everywhere in nature and in day-to-day life. To give a few examples, common fractals are the spiral cactus, the romanesco broccoli, human brain and the outline of the Swedish map. Fractal dimension is a dimension which need not take integer values. In fractal geometry, a fractal dimension is a ratio providing an index of the complexity of fractal pattern with regard to how the local geometry changes with the scale at which it is measured. In recent years, fractal analysis is used increasingly in many areas of engineering and technology. Among others, fractal analysis is used in signal and image compression, computer and video design, neuroscience and fractal based cancer modelling and diagnosing.   This study consists of two main parts. The first part of the study aims to understand the appearance of an irregular Cantor set generated by the chaotic dynamical system generated by the logistic function on the unit interval [0,1]. In order to understand this irregular Cantor set, we studied the topological properties of the Cantor Middle-thirds set and the generalised Cantor sets, all of which have zero length. The necessity to compare these sets with regard to their size led us to the second part of this paper, namely the dimension studies of fractals. More complex fractals were presented in the second part, three definitions of dimension were introduced. The fractal dimension of the irregular Cantor set generated by the logistic mapping was estimated and we found that the Hausdorff dimension has the widest scope and greatest flexibility in the fractal studies. / Fraktaler är ett relativt nytt ämne inom matematik som fick sitt stora genomslag först efter 60-talet.  En fraktal är ett självliknande mönster med struktur i alla skalor. Några vardagliga exempel på fraktaler är spiralkaktus, romanescobroccoli, mänskliga hjärnan, blodkärlen och Sveriges fastlandskust. Bråktalsdimension är en typ av dimension där dimensionsindexet tillåts att anta alla icke-negativa reella tal. Inom fraktalgeometri kan dimensionsindexet betraktas som ett komplexitetsindex av mönstret med avseende på hur den lokala geometrin förändras beroende på vilken skala mönstret betraktas i. Under det senaste decenniet har fraktalanalysen använts alltmer flitigt inom tekniska och vetenskapliga tillämpningar. Bland annat har fraktalanalysen använts i signal- och bildkompression, dator- och videoformgivning, neurovetenskap och fraktalbaserad cancerdiagnos.   Denna studie består av två huvuddelar. Den första delen fokuserar på att förstår hur en fraktal kan uppstå i ett kaotiskt dynamiskt system. För att vara mer specifik studerades den logistiska funktionen och hur denna ickelinjära avbildning genererar en oregelbunden Cantormängd på intervalet [0,1]. Vidare, för att förstå den oregelbundna Cantormängden studerades Cantormängden (eng. the Cantor Middle-Thirds set) och de generaliserade Cantormängderna, vilka alla har noll längd. För att kunna jämföra de olika Cantormängderna med avseende på storlek, leds denna studie vidare till dimensionsanalys av fraktaler som är huvudämnet i den andra delen av denna studie. Olika topologiska fraktaler presenterades, tre olika definitioner av dimension introducerades, bland annat lådräkningsdimensionen och Hausdorffdimensionen. Slutligen approximerades dimensionen av den oregelbundna Cantormängden med hjälp av Hausdorffdimensionen. Denna studie demonstrerar att Hausdorffdimensionen har större omfattning och mer flexibilitet för fraktalstudier.