1,720,992 research outputs found
Feature-basiertes Tracking mittels Bag of Visual Words
No full textMotivation dieser Arbeit ist die Entwicklung eines Programms zur Unterstützung von Monteuren bei der Wartung und Reparatur eines Fahrzeugs. Durch automatische Einblendungen, welche in der erfassten Szene angezeigt werden, wird vorgegeben welcher Schritt als nächstes ausgeführt werden soll. Unter anderem sollen die zu bearbeitenden Stellen durch das Einblenden von CAD-Objekten hervorgehoben werden. Um die Objekte an der richtigen Stelle einzublenden, muss die aktuelle Position und Orientierung der Kamera bekannt sein. Zur Berechnung der Kamerapose ist ein zuverlässiges Feature-Matching nötig, bei dem die benötigte 3D-Position der Features bestimmt wird. Da das System in Echtzeit arbeiten soll, kommt es hierbei vor allem auf eine schnelle aber dennoch genaue Ausführung an. Eine vielversprechende Methode zur Verkürzung der benötigten Rechenzeit besteht in der Kombination von Matching mit Visual Words. Das Ziel dieser Arbeit ist daher die Entwicklung eines Systems zum Feature-Matching mit Hilfe der Visual Words und wird ergänzt durch die anschließende Berechnung der Kamerapose. Als Feature-Detektor wird der SURF-Algorithmus verwendet. Die Visual Words werden umgesetzt, indem in der Lernphase die Features aller Bilder in Cluster eingeteilt werden. In Histogrammen wird festgehalten, wie sich die Features eines Bildes auf die Cluster verteilen. Beim späteren Tracking wird anhand von aktuellen Features ein Histogramm erstellt und mit den gespeicherten Histogrammen verglichen. Für das Matching werden nur diejenigen Bilder ausgewählt, die ein ähnliches Histogramm besitzen. Dies führt zu einer besseren und vor allem schnelleren Auswahl der Features, da Features, welche in diesem Bildausschnitt nicht vorkommen können, von der Suche ausgeschlossen werden. Dieses Vorgehen ermöglicht ein schnelles und gleichzeitig genaues Matching, wodurch auch die Berechnung der Kamerapose beschleunigt wird. Während die berechnete Kamerapose für ein durchgängiges Tracking möglicherweise nicht stabil genug ist, bietet sich das Verfahren vor allem für eine schnelle Initialisierungsphase an.This thesis is motivated by the idea of supporting mechanics with the repair and maintenance of vehicles. Automatic overlays of the examined scene guide the mechanics through each step as the relevant parts are highlighted by the overlay of CAD Objects. In order to display the objects at the right area of the image, the current position and orientation of the camera needs to be known. A reliable feature matching technique is required to calculate the 3D position of the camera. The system must work in real time, therefore a fast but accurate implementation is mandatory. A promising proceeding to reduce the calculation time results from the combination of the matching process with visual words. The intention of this thesis consists in the development of a feature based matching system which is supported by visual words. Finally, the position and orientation of the camera will be calculated. The proceeding uses the SURF-algorithm as a feature detector. During the learning phase the features of every image are assigned to clusters resulting in a histogram for each image. Later, during the tracking process, a new histogram, created from the identified features in real time, is compared to the stored histogram. Only certain images which lead to a similar histogram are selected for the matching process which results in a more accurate and faster selection. This improvement follows from eliminating several features which aren't contained within the examined area of the scene. The discussed approach allows for a fast and accurate matching process which also results in an accelerated calculation of the camera pose. As the results aren't quite stable enough for a continuous tracking, the proceeding is still suitable for a quick initialization process
Modifizierter ICP Algorithmus zum Tracking von CAD Geometrien mithilfe von Tiefendaten
No full textAs both cameras and processors have improved massively over the years, the application range for object tracking algorithms has grown as well, now reaching from augmented reality over visual perception in robotics to very precise real-time visual quality control in the industry. However, the current approaches around RGB tracking lack robustness for bad lighting conditions or models without strong and distinct edges which creates a need for algorithms on depth data. Most current depth tracking algorithms are based on the Iterative Closest Point algorithm to align two depth measurements. So far, ICP variants have strictly kept the creation of correspondences and the minimization of the global error over these correspondences separate. The research goal of this work is to design and implement a modified ICP frame-to-model tracking pipeline with integrated Projective Data Association in the optimization loop for additional efficiency. After that, we enhance the normal vector computation, interpolation techniques for the Projective Data Association, and outlier removal strategies. By testing on synthetic data, we can show that these modular refinements improve the overall tracking performance in terms of precision, robustness, convergence radius, and efficiency. The total pipeline reaches a convergence rate of around 30% on the synthetic data in our test setup. While this does not suffice for applications in the real world yet, it still shows that the approach looks promising and bears potential for future research on further enhancements
Optimal Spatial Registration of SLAM for Augmented Reality
No full textAugmented reality (AR) is a paradigm that aims at fusing the perceived real environment of a human with digital information located in 3D space. Typically, virtual 3D graphics are overlayed into the captured images of a moving camera or directly into the user's field-of-view by means of optical see-through displays (OST). For a correct perspective and view-dependent alignment of the visualization, it is required to solve various static and dynamic geometric registration problems in order to create the impression that the virtual and the real world are seamlessly interconnected.
The advances during the last decade in the field of simultaneous localization and mapping (SLAM) represent an important contribution to this general problem. It is now possible to reconstruct the real environment and to simultaneously capture the dynamic movements of a camera from the images without having to instrument the environment in advance. However, SLAM in general can only partly solve the entire registration problem, because the retrieved 3D scene geometry and the calculated motion path are spatially related only with regard to an arbitrarily selected coordinate system. Without a proper reconciliation of coordinate systems (spatial registration), the real world of the human observer still remains decoupled from the virtual world. Existing approaches for solving this problem either require the availability of a virtual 3D model that represents a real object with sufficient accuracy (model-based tracking), or they rely on use-case specific assumptions and additional sensor data (such as GPS signals or the Manhattan-world assumption). Therefore, these approaches are bound to these additional prerequisites, which limit the general applicability. The circumstance that automated registration is desirable but not always possible, creates the need for techniques that allow a user to specify connections between the real and the virtual world when setting up AR applications, so that it becomes possible to support and control the process of registration. These techniques must be complemented with numerical algorithms that optimally exploit the provided information to obtain precise registration results.
Within this context, the present thesis provides the following contributions.
* We propose a novel, closed-form (non-iterative) algorithm for calculating a Euclidean or a similarity transformation. The presented algorithm is a generalization of recent state-of-the-art solvers for computing the camera pose based on 2D measurement points in the image (perspective-n-point problem) - a fundamental problem in computer vision that has attracted research for many decades. The generalization consists in extending and unifying these algorithms, so that they can handle other types of input correspondences than originally designed for. With this algorithm, it becomes possible to perform a rigid registration of SLAM systems to a target coordinate system based on heterogeneous and partially indeterminate input data.
* We address the global refinement of structure and motion parameters by means of iterative sparse minimization (bundle adjustment or BA), which has become a standard technique inside SLAM systems. We propose a variant of BA in which information about the virtual domain is integrated as constraints by means of an optimization-on-manifold approach. This serves for compensating low-frequency deformations (non-rigid registration) of the estimated camera path and the reconstructed scene geometry caused by measurement error accumulation and the ill-conditionedness of the BA problem.
* We present two approaches in which a user can contribute with his knowledge for registering a SLAM system. In a first variant, the user can place markers in the real environment with predefined connections to the virtual coordinate system. Precise positioning of the markers is not required, rather they can be placed arbitrarily on surfaces or along edges, which notably facilitates the preparative effort. During run-time, the dispersed information is collected and registration is accomplished automatically. In a second variant, the user is given the possibility to mark salient points in an image sequence during a preparative preprocessing step and to assign corresponding points in the virtual 3D space via a simple point-and-click metaphor. The result of this preparative phase is a precisely registered and ready-to-use reference model for camera tracking at run-time.
* Finally, we propose an approach for geometric calibration of optical see-trough displays. We present a parametric model, which allows to dynamically adapt the rendering of virtual 3D content to the current viewpoint of the human observer, including a pre-correction of image aberrations caused by the optics or irregularly curved combiners. In order to retrieve its parameters, we propose a camera-based approach, in which elements of the real and the virtual domain are simultaneously observed. The calibration procedure was developed for a head-up display in a vehicle. A prototypical extension to head-mounted displays is also presented.EG Graphics Dissertation Onlin
Optimal Spatial Registration of SLAM for Augmented Reality
Get PDFAugmented reality (AR) is a paradigm that aims at fusing the perceived real environment of a human with digital information located in 3D space. Typically, virtual 3D graphics are overlayed into the captured images of a moving camera or directly into the user's field-of-view by means of optical see-through displays (OST). For a correct perspective and view-dependent alignment of the visualization, it is required to solve various static and dynamic geometric registration problems in order to create the impression that the virtual and the real world are seamlessly interconnected.
The advances during the last decade in the field of simultaneous localization and mapping (SLAM) represent an important contribution to this general problem. It is now possible to reconstruct the real environment and to simultaneously capture the dynamic movements of a camera from the images without having to instrument the environment in advance. However, SLAM in general can only partly solve the entire registration problem, because the retrieved 3D scene geometry and the calculated motion path are spatially related only with regard to an arbitrarily selected coordinate system. Without a proper reconciliation of coordinate systems (spatial registration), the real world of the human observer still remains decoupled from the virtual world. Existing approaches for solving this problem either require the availability of a virtual 3D model that represents a real object with sufficient accuracy (model-based tracking), or they rely on use-case specific assumptions and additional sensor data (such as GPS signals or the Manhattan-world assumption). Therefore, these approaches are bound to these additional prerequisites, which limit the general applicability. The circumstance that automated registration is desirable but not always possible, creates the need for techniques that allow a user to specify connections between the real and the virtual world when setting up AR applications, so that it becomes possible to support and control the process of registration. These techniques must be complemented with numerical algorithms that optimally exploit the provided information to obtain precise registration results.
Within this context, the present thesis provides the following contributions.
* We propose a novel, closed-form (non-iterative) algorithm for calculating a Euclidean or a similarity transformation. The presented algorithm is a generalization of recent state-of-the-art solvers for computing the camera pose based on 2D measurement points in the image (perspective-n-point problem) - a fundamental problem in computer vision that has attracted research for many decades. The generalization consists in extending and unifying these algorithms, so that they can handle other types of input correspondences than originally designed for. With this algorithm, it becomes possible to perform a rigid registration of SLAM systems to a target coordinate system based on heterogeneous and partially indeterminate input data.
* We address the global refinement of structure and motion parameters by means of iterative sparse minimization (bundle adjustment or BA), which has become a standard technique inside SLAM systems. We propose a variant of BA in which information about the virtual domain is integrated as constraints by means of an optimization-on-manifold approach. This serves for compensating low-frequency deformations (non-rigid registration) of the estimated camera path and the reconstructed scene geometry caused by measurement error accumulation and the ill-conditionedness of the BA problem.
* We present two approaches in which a user can contribute with his knowledge for registering a SLAM system. In a first variant, the user can place markers in the real environment with predefined connections to the virtual coordinate system. Precise positioning of the markers is not required, rather they can be placed arbitrarily on surfaces or along edges, which notably facilitates the preparative effort. During run-time, the dispersed information is collected and registration is accomplished automatically. In a second variant, the user is given the possibility to mark salient points in an image sequence during a preparative preprocessing step and to assign corresponding points in the virtual 3D space via a simple point-and-click metaphor. The result of this preparative phase is a precisely registered and ready-to-use reference model for camera tracking at run-time.
* Finally, we propose an approach for geometric calibration of optical see-trough displays. We present a parametric model, which allows to dynamically adapt the rendering of virtual 3D content to the current viewpoint of the human observer, including a pre-correction of image aberrations caused by the optics or irregularly curved combiners. In order to retrieve its parameters, we propose a camera-based approach, in which elements of the real and the virtual domain are simultaneously observed. The calibration procedure was developed for a head-up display in a vehicle. A prototypical extension to head-mounted displays is also presented
Schlussbericht zum Vorhaben im Rahmen des Eurostars Projekts E!601 HERON - "Kognitiv unterstützte mobile Robotik für flexible kollaborative Montagearbeiten"
No full textOptimal Spatial Registration of SLAM for Augmented Reality
No full textAugmented reality (AR) is a paradigm that aims at fusing the perceived real environment of a human with digital information located in 3D space. Typically, virtual 3D graphics are overlayed into the captured images of a moving camera or directly into the user's field-of-view by means of optical see-through displays (OST). For a correct perspective and view-dependent alignment of the visualization, it is required to solve various static and dynamic geometric registration problems in order to create the impression that the virtual and the real world are seamlessly interconnected. The advances during the last decade in the field of simultaneous localization and mapping (SLAM) represent an important contribution to this general problem. It is now possible to reconstruct the real environment and to simultaneously capture the dynamic movements of a camera from the images without having to instrument the environment in advance. However, SLAM in general can only partly solve the entire registration problem, because the retrieved 3D scene geometry and the calculated motion path are spatially related only with regard to an arbitrarily selected coordinate system. Without a proper reconciliation of coordinate systems (spatial registration), the real world of the human observer still remains decoupled from the virtual world. Existing approaches for solving this problem either require the availability of a virtual 3D model that represents a real object with sufficient accuracy (model-based tracking), or they rely on use-case specific assumptions and additional sensor data (such as GPS signals or the Manhattan-world assumption). Therefore, these approaches are bound to these additional prerequisites, which limit the general applicability. The circumstance that automated registration is desirable but not always possible, creates the need for techniques that allow a user to specify connections between the real and the virtual world when setting up AR applications, so that it becomes possible to support and control the process of registration. These techniques must be complemented with numerical algorithms that optimally exploit the provided information to obtain precise registration results. Within this context, the present thesis provides the following contributions.* We propose a novel, closed-form (non-iterative) algorithm for calculating a Euclidean or a similarity transformation. The presented algorithm is a generalization of recent state-of-the-art solvers for computing the camera pose based on 2D measurement points in the image (perspective-n-point problem) - a fundamental problem in computer vision that has attracted research for many decades. The generalization consists in extending and unifying these algorithms, so that they can handle other types of input correspondences than originally designed for. With this algorithm, it becomes possible to perform a rigid registration of SLAM systems to a target coordinate system based on heterogeneous and partially indeterminate input data.* We address the global refinement of structure and motion parameters by means of iterative sparse minimization (bundle adjustment or BA), which has become a standard technique inside SLAM systems. We propose a variant of BA in which information about the virtual domain is integrated as constraints by means of an optimization-on-manifold approach. This serves for compensating low-frequency deformations (non-rigid registration) of the estimated camera path and the reconstructed scene geometry caused by measurement error accumulation and the ill-conditionedness of the BA problem.* We present two approaches in which a user can contribute with his knowledge for registering a SLAM system. In a first variant, the user can place markers in the real environment with predefined connections to the virtual coordinate system. Precise positioning of the markers is not required, rather they can be placed arbitrarily on surfaces or along edges, which notably facilitates the preparative effort. During run-time, the dispersed information is collected and registration is accomplished automatically. In a second variant, the user is given the possibility to mark salient points in an image sequence during a preparative preprocessing step and to assign corresponding points in the virtual 3D space via a simple point-and-click metaphor. The result of this preparative phase is a precisely registered and ready-to-use reference model for camera tracking at run-time.* Finally, we propose an approach for geometric calibration of optical see-trough displays. We present a parametric model, which allows to dynamically adapt the rendering of virtual 3D content to the current viewpoint of the human observer, including a pre-correction of image aberrations caused by the optics or irregularly curved combiners. In order to retrieve its parameters, we propose a camera-based approach, in which elements of the real and the virtual domain are simultaneously observed. The calibration procedure was developed for a head-up display in a vehicle. A prototypical extension to head-mounted displays is also presented.Erweiterte Realität (AR) bezeichnet ein Paradigma, welches darauf abzielt, die wahrgenommene, reale Umgebung eines Menschen mit im 3D-Raum verorteten, digitalen Informationen zu verschmelzen. Typischerweise werden dabei in Echtzeit virtuelle 3D Grafiken in die aufgenommenen Bilder einer sich bewegenden Kamera oder direkt in das Sichtfeld des Nutzers über optische Durchsichtanzeigen eingebettet. Die perspektivisch korrekte und lagerichtige Darstellung erfordert hierzu die Lösung verschiedener statischer und dynamischer Registrierungsprobleme, um den Eindruck zu erzeugen, dass die virtuelle und die reale Welt nahtlos miteinander verbunden sind. Die im letzten Jahrzehnt erreichten Fortschritte im Bereich simultaner Lokalisierung und Kartierung (SLAM) liefern hierzu einen wichtigen Beitrag. Hiermit ist es möglich, die reale Umgebung zu rekonstruieren und dabei gleichzeitig die dynamische Eigenbewegung einer Kamera aus den Bildern zu erfassen, ohne dass die Umgebung hierzu präpariert werden muss. Dennoch löst SLAM damit nur einen Teil des gesamten Registrierungsproblems, da die erstellte 3D Szenengeometrie und der berechnete Bewegungspfad räumlich nur in Bezug zu einem frei gewählten Koordinatensystem gesetzt wird. Ohne einen entsprechenden Abgleich der Koordinatensysteme bleibt die reale Welt des menschlichen Beobachters stets von der virtuellen Welt entkoppelt. Bestehende Ansätze zur Lösung dieses Problems erfordern entweder die Verfügbarkeit eines virtuellen 3D-Modells, welches einem realen Objekt mit ausreichender Genauigkeit entsprechen muss (modellbasiertes Tracking), oder sie stützen sich auf anwendungsfallspezifische Annahmen und zusätzliche Sensordaten (wie GPS-Signale oder der "Manhattan Welt"-Annahme), welches die allgemeine Anwendbarkeit dieser Verfahren einschränkt. Der Umstand, dass eine automatisierte Registrierung wünschenswert jedoch nicht immer möglich ist, schafft den Bedarf an Techniken, mit denen ein Benutzer beim Einrichten von AR-Anwendungen Verbindungen zwischen der realen und der virtuellen Welt spezifizieren und somit die Registrierung begleiten und kontrollieren kann. Diese Techniken benötigen die Unterstützung durch numerische Algorithmen, welche die Informationen optimal ausnutzen, um somit genaue Registrierungsergebnisse zu erreichen. In diesem Zusammenhang liefert die vorliegende Arbeit die folgenden Beiträge. * Es wird ein neuartiger, nicht-iterativer Algorithmus zur Berechnung einer euklidischen Transformation oder einer Ähnlichkeitstransformation präsentiert. Der vorgestellte Algorithmus stellt eine Verallgemeinerung neuester Ansätze zur Berechnung der Kameraposition und -Orientierung anhand von 2D-Messpunkten im Bild dar (räumlicher Rückwärtsschnitt) - ein grundlegendes Problem im Bereich des maschinellen Sehens mit einer langen Forschungshistorie. Die Verallgemeinerung besteht darin, diese Algorithmen so zu erweitern und zu vereinheitlichen, dass sie mit anderen Arten von Eingangskorrespondenzen als ursprünglich vorgesehen umgehen können. Der Algorithmus ermöglicht es, eine rigide Registrierung von SLAM-Systemen zu einem Zielkoordinatensystem auf der Grundlage heterogener und partiell unbestimmter Eingangsdaten durchzuführen. * Zudem wird die globale Minimierung von Struktur- und Bewegungsparametern durch den Bündelblockausgleich (BA) addressiert, welcher sich als Standardtechnik innerhalb von SLAM-Systemen etabliert hat. In dieser Arbeit wird hierzu eine Variante des BA vorgeschlagen, bei der Informationen über die virtuelle Domäne als Gleichheitsnebenbedingungen integriert werden, wobei hierfür als Technik eine Parameteroptimierung entlang ihrer zugehörigen Mannigfaltigkeiten gewählt wird. Die Integration der Nebenbedigungen erfolgt mit dem Ziel, niederfrequente Deformationen (nicht-rigide Registrierung) des geschätzten Kamerapfades und der rekonstruierten Szenengeometrie zu kompensieren, welche durch die Messfehlerakkumulation und die schlechte Konditioniertheit des BA-Minmierungsproblems verursacht werden. * Weiterhin werden zwei Ansätze vorgestellt, bei denen ein Benutzer mit seinem Wissen zur Registrierung eines SLAM-Systems beitragen kann. In einer ersten Variante kann der Benutzer in der realen Umgebung Marker mit vordefinierten Verbindungen zum virtuellen Koordinatensystem platzieren. Eine genaue Positionierung der Marker ist nicht erforderlich, vielmehr können sie beliebig auf Oberflächen oder entlang von Kanten platziert werden, was den präparativen Aufwand erheblich reduziert. Zur Laufzeit werden dann die verteilten Informationen erfasst und die Registrierung des SLAM erfolgt automatisch. In einer zweiten Variante hat der Benutzer im Rahmen eines Vorverarbeitungsschrittes die Möglichkeit, markante Punkte in einer bereits aufgenommenen Bildsequenz zu selektieren und entsprechenden Punkten im virtuellen 3D-Raum über eine einfache Nutzerschnittstelle zuzuordnen. Das Ergebnis dieser Vorbereitung ist ein präzise registriertes Referenzmodel, welches zur Laufzeit unmittelbar zur Kameraverortung eingesetzt werden kann. * Schließlich wird ein Ansatz zur geometrischen Kalibrierung optischer Durchsichtanzeigen vorgeschlagen. Es wird ein parametrisches Modell vorgestellt, mit dem das Rendern von virtuellem 3D-Inhalt an den aktuellen Blickpunkt des menschlichen Beobachters dynamisch angepasst werden kann, einschließlich einer Vorkorrektur der durch die Optik oder unregelmäßig gekrümmter Kombinatoren verursachten Bildverzerrungen. Um die Modellparameter zu ermitteln, wird ein kamerabasierter Ansatz vorgeschlagen, bei dem Elemente der realen und der virtuellen Domäne gleichzeitig erfasst werden. Das Kalibrierungsverfahren wurde examplarisch für ein Head-up-Display in einem Fahrzeug entwickelt. Eine prototypische Erweiterung für Head-Mounted-Displays wird ebenfalls vorgestellt
Unifying algebraic solvers for scaled euclidean registration from point, line and plane constraints
No full textS.52-66We investigate recent state-of-the-art algorithms for absolute pose problems (PnP and GPnP) and analyse their applicability to a more general type, namely the scaled Euclidean registration from point-to-point, point-to-line and point-to plane correspondences. Similar to previous formulations we first compress the original set of equations to a least squares error function that only depends on the non-linear rotation parameters and a small symmetric coefficient matrix of fixed size. Then, in a second step the rotation is solved with algorithms which are derived using methods from algebraic geometry such as the Gröbner basis method. In previous approaches the first compression step was usually tailored to a specific correspondence types and problem instances. Here, we propose a unified formulation based on a representation with orthogonal complements which allows to combine different types of constraints elegantly in one single framework. We show that with our unified formulation existing polynomial solvers can be interchangeably applied to problem instances other than those they were originally proposed for. It becomes possible to compare them on various registrations problems with respect to accuracy, numerical stability, and computational speed. Our compression procedure not only preserves linear complexity, it is even faster than previous formulations. For the second step we also derive an own algebraic equation solver, which can additionally handle the registration from 3D point-to-point correspondences, where other solvers surprisingly fail
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
- …
