Το παρών έγγραφο παρουσιάζει μεθόδους για την τρισδιάστατη αναδημιουργία μιας άκαμπτης σκηνής ή αντικειμένου από δυο εικόνες. Δυο εικόνες μιας σκηνής σχετίζονται με την επιπολική γεωμετρία, η οποία μπορεί να περιγραφεί από μια 3x3 μήτρα η οποία καλείται θεμελιώδης μήτρα. Αρχικά, περιγράφεται ένας αλγόριθμος για την εύρεση αντιστοιχιών σημείων στις δυο εικόνες και στη συνέχεια μέθοδοι για την εκτίμηση της θεμελιώδους μήτρας από τις αντιστοιχίες αυτές. Μια προβολική αναδημιουργία της σκηνής λαμβάνεται από τη θεμελιώδη μήτρα, η οποία στην συνέχεια μπορεί να αναβαθμιστεί σε μια συσχετισμένη και τέλος σε μια μετρική αναδημιουργία χρησιμοποιώντας πρόσθετες πληροφορίες από την σκηνή. Επιπλέον, παρουσιάζεται μια εναλλακτική προσέγγιση, στην οποία εφαρμόζουμε μια μέθοδο αυτοβαθμονόμησης των καμερών από τις οποίες λήφθησαν οι εικόνες χωρίς τη χρήση ενός ειδικού αντικειμένου βαθμονόμησης και χωρίς καμία υπόθεση όσον αφορά τις παραμέτρους τους. Έπειτα, λαμβάνεται μια μετρική αναδημιουργία, η οποία στις περισότερες εφαρμογές είναι το ζητούμενο, μέσω της ουσιώδους μήτρας που είναι η αντίστοιχη της θεμελιώδους στην περίπτωση όπου οι εσωτερικές παράμετροι των καμερών είναι γνωστές. Τέλος, περιγράφονται διάφορες μέθοδοι με τις οποίες μπορούμε να υπολογίσουμε τη δομή της σκηνής, που αποτελεί το τελικό στάδιο της αναδημιουργίας. Σημειώνουμε ότι μεγαλύτερη έμφαση έχει δοθεί σε μεθόδους οι οποίες είναι ανθεκτικές στο θόρυβο, δεδομένης της παρουσίας του κατά τη λήψη των εικόνων, ώστε η λαμβανόμενη αναδημιουργία να είναι αξιόπιστη και ικανοποιητικής ακρίβειας. / This paper describes methods for the 3d reconstruction of a scene from two images. The epipolic geometry of two images can be described by a 3x3 matric, called the funtamental matrix. Firstly, an algorithm is described that finds corresponding points in thw two images and then thiw corresponding points are used for the estimation of the funtamemtal matrix. A prolective reconstruction of a scene is obtained from the funtamental matrix. This reconstruction will then be upgraided to an affine one, and finally to a metric, using additional information from the scene. Finally, we describe a number of methods for the triangulation of the scene. We denote that because of the noise in the data, we have emphasise at robust methods

Les pratiques de conservation et restauration de monuments historiques requièrent l’élaboration de diagnostics impliquant différents intervenants au sein de contextes d’études pluridisciplinaires. L’état de conservation d’un objet patrimonial est ainsi étudié et décrit au moyen d’observations directes, de sources documentaires et de données analytiques de natures différentes. Les avancées des technologies numériques en matière de collecte, traitement et gestion de données, offrent aujourd’hui une opportunité sans précédent pour intégrer les résultats de ces observations et ces données au sein de systèmes innovants de représentation pour la documentation et la connaissance du patrimoine. Cependant, si une panoplie de nouveaux outils est aujourd’hui à disposition de la communauté des scientifiques et des professionnels du patrimoine, le problème d’une corrélation pertinente de ces données et de ces informations hétérogènes reste peu exploré. Tout d’abord, si ces nouveaux outils permettent aux différents experts de mémoriser et d’analyser leurs observations sur différents supports, les données générées par ces différents experts ne sont généralement pas spatialisées autour d’un même référentiel spatial. En effet, même si toutes ces données se réfèrent à un objet physique commun, les liens entre elles ne peuvent que s’établir par des stratégies d’organisation de fichiers ou par des méthodes d’indexation basées sur des mots clés. Parallèlement, si dans les dernières années plusieurs techniques de numérisation 3D ont été expérimentées dans le but de générer des représentations géométriques denses et précises, les méthodes de traitement et de structuration de ces données 3D ne fournissent pas encore de cadres opérationnels pour l’extraction d'informations pertinentes pour l’analyse et l’interprétation de l’état de conservation. En se positionnant à l'intersection entre les domaines de l’acquisition spatialisée 3D et des systèmes d’informations, ce travail de recherche propose une ontologie de domaine dédiée à l’annotation sémantique de représentations 3D d’objets patrimoniaux visant à la constitution d’un environnement numérique pour la description de l’état de conservation des monuments historiques. Par l’interconnexion de descripteurs qualitatifs (reliés à une formalisation des connaissances du domaine) et qualitatifs, cette ontologie constitue l'échafaudage conceptuel structurant un système d’informations multidimensionnelles dédié à la corrélation spatiale, géométrique et sémantique de jeux d’annotations élaborés par des acteurs multiples et en fonction de niveaux de lecture multiples. / The conservation and restauration of historical monuments require a diagnostic analysis carried out by a multidisciplinary team. The elaboration of the diagnosis of a cultural Heritage Object requires direct observations, the examination of documentary sources as well as of diverse types of analytic data. The great advancements in digital processing, management and data collection opened unprecedented opportunities for integrating results, coming from both observations and derivative data, within innovating representation systems for heritage knowledge and documentation. However, if a range of new tools and data is today available to the scientific community and heritage experts, their correlation and integration with internal/external heterogeneous information is an issue that still remain unexplored. While these innovative tools allow different experts to record and analyze their observations with diverse formats, the results are generally not spatialized and referenced together. Indeed, even though all these data refer to a common physical object, the links between them is based only on file organization strategies or by keyword-based indexation methods. In parallel, in recent years, several 3D digitization technics has been used for generating dense and accurate geometrical representations, but the processing and structuration method of these 3D data don’t include yet an operational framework for retrieving relevant information regarding their conservation state and an interpretative analysis. Positioned in the intersection between 3D spatialized acquisition domain and information management, this research work aims to the creation of a digital framework for recording conservation state description of historical monument throughout the introduction of a functional domain ontology for the semantic annotations of heritage objects 3D representations. The proposed ontology comprises both qualitative (related to a domain knowledge formalization) and quantitative descriptors, constituting the necessary conceptual scaffold for structuring a multidimensional information system dedicated to the correlation of spatial, geometrical and semantic multi-actor annotations in relation to multiple observation levels.

De nombreux domaines d'applications utilisent des représentations spatiales. C'est le cas de l'architecture, de l'urbanisme ou de la géographie. L'acquisition de ces données spatiales en urbanisme a connu ces dernières années des progrès significatifs grâce à l'introduction de nouveaux instruments de mesure. Ces derniers permettent d'obtenir des supports d'analyse urbaine à différents niveaux de détails et pour différentes finalités. Cette thèse propose une approche permettant de combiner deux disciplines, la typo-morphologie urbaine et la géomatique. Nous avons expliqué la notion centrale de processus morphologique, les différentes étapes d'opération propre à l'analyse historique, notamment le traitement des sources cartographiques avec l'outil S.I.G. Notre travail a avant tout consisté à explorer l'apport des S.I.G. au traitement et à l'analyse des données historiques. Nous nous sommes intéressés particulièrement à l'approche typo-morphologique pour compléter le potentiel interprétatif et descriptif. Notre travail de thèse est issu de différentes étapes, nous pouvons mentionner la construction d'une classification formelle, des concepts liés à l'évolution historique et morphologique des villes de Constantine et d'Aix-en-Provence. En partant de cette histoire urbaine, comparer ces deux villes à permis d'établir une chronologie de l'évolution des formes urbaines, de mieux saisir les enjeux de chacune de celles-ci. / Many fields of applications use spatial representations. This is the case of architecture, town planning or geography. The acquisition of these spatial datas in town planning these last years has experienced a significant progress with the introduction of new instruments. This acquisition allows to get urban support analysis at different levels of details and for different purposes. This thesis proposes an approach to combine two disciplines, the urban typomorphology and geomatics. We have explained the central notion of morphological process, the different steps of operation peculiar to the historical analysis for the treatment of map datas with the GIS instrument, primarily our work consist to explore the GIS contribution to the historical data treatment and analysis. We focused particularly on the approach to complete typomorphological potential interpretive and descriptive. Our thesis work has been made from different stages, we can mention the construction of a formal classification, concepts related to the historical development and morphology of Constantine and Aix-en-Provence. Starting from this urban history compare the two cities has established a chronology of the evolution of urban forms, to better understand the challenges each of these latter. Specifically, this work allows us to contribute to improving the mastery of the urban project. Finally tracks are proposed to continue this work by exploiting the platform exploration of 3D representation proved very useful for making historical analysis.

The research described in this thesis was motivated by the need of a robust model capable of representing 3D data obtained with 3D sensors, which are inherently noisy. In addition, time constraints have to be considered as these sensors are capable of providing a 3D data stream in real time. This thesis proposed the use of Self-Organizing Maps (SOMs) as a 3D representation model. In particular, we proposed the use of the Growing Neural Gas (GNG) network, which has been successfully used for clustering, pattern recognition and topology representation of multi-dimensional data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models, without considering time constraints. It is proposed a hardware implementation leveraging the computing power of modern GPUs, which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). The proposed methods were applied to different problem and applications in the area of computer vision such as the recognition and localization of objects, visual surveillance or 3D reconstruction.

Nowadays, new computers generation provides a high performance that enables to build computationally expensive computer vision applications applied to mobile robotics. Building a map of the environment is a common task of a robot and is an essential part to allow the robots to move through these environments. Traditionally, mobile robots used a combination of several sensors from different technologies. Lasers, sonars and contact sensors have been typically used in any mobile robotic architecture, however color cameras are an important sensor due to we want the robots to use the same information that humans to sense and move through the different environments. Color cameras are cheap and flexible but a lot of work need to be done to give robots enough visual understanding of the scenes. Computer vision algorithms are computational complex problems but nowadays robots have access to different and powerful architectures that can be used for mobile robotics purposes. The advent of low-cost RGB-D sensors like Microsoft Kinect which provide 3D colored point clouds at high frame rates made the computer vision even more relevant in the mobile robotics field. The combination of visual and 3D data allows the systems to use both computer vision and 3D processing and therefore to be aware of more details of the surrounding environment. The research described in this thesis was motivated by the need of scene mapping. Being aware of the surrounding environment is a key feature in many mobile robotics applications from simple robotic navigation to complex surveillance applications. In addition, the acquisition of a 3D model of the scenes is useful in many areas as video games scene modeling where well-known places are reconstructed and added to game systems or advertising where once you get the 3D model of one room the system can add furniture pieces using augmented reality techniques. In this thesis we perform an experimental study of the state-of-the-art registration methods to find which one fits better to our scene mapping purposes. Different methods are tested and analyzed on different scene distributions of visual and geometry appearance. In addition, this thesis proposes two methods for 3d data compression and representation of 3D maps. Our 3D representation proposal is based on the use of Growing Neural Gas (GNG) method. This Self-Organizing Maps (SOMs) has been successfully used for clustering, pattern recognition and topology representation of various kind of data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models without considering time constraints. Self-organising neural models have the ability to provide a good representation of the input space. In particular, the Growing Neural Gas (GNG) is a suitable model because of its flexibility, rapid adaptation and excellent quality of representation. However, this type of learning is time consuming, specially for high-dimensional input data. Since real applications often work under time constraints, it is necessary to adapt the learning process in order to complete it in a predefined time. This thesis proposes a hardware implementation leveraging the computing power of modern GPUs which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). Our proposed geometrical 3D compression method seeks to reduce the 3D information using plane detection as basic structure to compress the data. This is due to our target environments are man-made and therefore there are a lot of points that belong to a plane surface. Our proposed method is able to get good compression results in those man-made scenarios. The detected and compressed planes can be also used in other applications as surface reconstruction or plane-based registration algorithms. Finally, we have also demonstrated the goodness of the GPU technologies getting a high performance implementation of a CAD/CAM common technique called Virtual Digitizing.

This thesis considers atom`s attributes, which affects shape of molecules. It describes rules that are basic for molecule creation. This text features basic attributes and rules, which affects the final shape of molecule. Next part of this text explains several ways to display molecule. Following parts of this thesis describes several ways how to store and represent atom and molecules in computer technology along with solution of these problems that are used for this thesis.