Refine
Year of publication
Document Type
- Conference Proceeding (90)
- Book (21)
- Article (peer reviewed) (11)
- Other (10)
- Contribution to a Periodical (9)
- Part of a Book (4)
- Doctoral Thesis (2)
- Working Paper (1)
Is part of the Bibliography
- no (148)
Keywords
- Speech Recognition (18)
- CFT (6)
- Software Development (6)
- Database (5)
- Microsoft (5)
- Sql server (5)
- Augmented Reality (AR) (4)
- Fault trees (4)
- Mobile Robot (4)
- Robotic Research (4)
Institute
- Fakultät für Informatik (148) (remove)
Reverberation effects as observed by room microphones severely degrade the performance of automatic speech recognition systems. We investigate the use of dereverberation by spectral subtraction as proposed by Lebart and Boucher and introduce a simple approach to estimate the required decay parameter by clapping hands. Experiments on small vocabulary continuous speech recognition task on read speech show that using the calibrated dereverberation improves WER from 73.2 to 54.7 for the best microphone. In combination with system adaptation, the WER could be reduced to 28.2, which is only a 16% relative loss of performance comparison to using a headset instead of a room microphone.
This contribution introduces MOBSY, a fully integrated, autonomous mobile service robot system. It acts as an automatic dialogue-based receptionist for visitors to our institute.
MOBSY incorporates many techniques from different research areas into one working stand-alone system. The techniques involved range from computer vision over speech understanding to classical robotics.
Along with the two main aspects of vision and speech, we also focus on the integration aspect, both on the methodological and on the technical level.
We describe the task and the techniques involved. Finally, we discuss the experiences that we gained with MOBSY during a live performance at our institute.
Die typischerweise gewünschten Einsatzgebiete für Dienstleistungsroboter, z. B. Krankenhäuser oder Seniorenheime, stellen sehr hohe Anforderungen an die Mensch-Maschine-Schnittstelle.
Diese Erfordernisse gehen im Allgemeinen über die Möglichkeiten der Standardsensoren, wie Ultraschalloder Infrarotsensoren, hinaus. Es müssen daher ergänzende Verfahren zum Einsatz kommen.
Aus der Sicht der Mustererkennung sind die Nutzung des Rechnersehens und des natürlichsprachlichen Dialogs von besonderem Interesse. Dieser Beitrag stellt das mobile System MOBS Y vor. MOBS Y ist ein vollkommen integrierter autonomer mobiler Dienstleistungsroboter.
Er dient als ein automatischer dialogbasierter Empfangsservice für Besucher unseres Instituts.
MOBSY vereinigt vielfältige Methoden aus unterschiedlichsten Forschungsgebieten in einem eigenständigen System. Die zum Einsatz kommenden Methoden aus dem Bereich der Bildverarbeitung reichen dabei von Objektklassifikation über visuelle Selbstlokalisierung und Rekalibrierung bis hin zu multiokularer Objektverfolgung.
Die Dialogkomponente umfasst Methoden der Spracherkennung, des Sprachverstehens und die Generierung von Antworten. Im Beitrag werden die zu erfüllende Aufgabe und die einzelnen Verfahren dargestellt.
MOBSY is a fully integrated autonomous mobile service robot system.
It acts as an automatic dialogue based receptionist for visitors of our institute. MOBSY incorporates many techniques from different research areas into one working stand-alone system. Especially the computer vision and dialogue aspects are of main interest from the pattern recognition’s point of view.
To summarize shortly, the involved techniques range from object classification over visual self-localization and recalibration to object tracking with multiple cameras. A dialogue component has to deal with speech recognition, understanding and answer generation. Further techniques needed are navigation, obstacle avoidance, and mechanisms to provide fault tolerant behavior.
This contribution introduces our mobile system MOBSY. Among the main aspects vision and speech, we focus also on the integration aspect, both on the methodological and on the technical level. We describe the task and the involved techniques.
Finally, we discuss the experiences that we gained with MOBSY during a live performance at the 25th anniversary of our institute
Wepresentaniterativeregistrationalgorithmfor aligning two differently scaled 3-D point sets. It extends the popular Iterative Closest Point (ICP) algorithm by estimating a scale factor between the two point sets in every iteration.
The presented algorithm is especially useful for the registration of point sets generated by structure-frommotion algorithms, which only reconstruct the 3-D structure of a scene upto scale. LiketheoriginalICPalgorithm,thepresentedalgorithm requires a rough pre-alignment of the point sets.
In order to determine the necessary accuracy of the pre-alignment, wehaveexperimentallyevaluatedthebasinofconvergence of the algorithm with respect to the initial rotation, translation, andscale factor between the two point sets.
There are many nearest neighbor algorithms tailor made for ICP,but most of them require Special input data like range Images or triangle meshes.
We focus on efficient nearest neighbor algorithms that do not impose this limitation, and thus can also be used with 3-D point sets generated by structure-frommotion techniques. We shortly present the evaluated algorithms and introduce the modifications we made to improve their efficiency.
In particular, several enhancements to the well-known k-D tree algorithm are described. The first part of our Performance Analysis consists of Experiments on synthetic point sets, whereas the second part features experiments with the ICP algorithm on real point sets. Both parts are completed by a thorough evaluation of the obtained results.
Robust registration of two 3-D point sets is a common problem in computer vision.
The iterative closest point (ICP) algorithm is undoubtedly the most popular algorithm for solving this kind of problem. In this paper, we present the Picky ICP algorithm, which has been created by merging several extensions of the standard ICP algorithm, thus improving its robustness and computation time.
Using pure 3-D point sets as input data, we do not consider additional information like point color or neighborhood relations. In addition to the standard ICP algorithm and the Picky ICP algorithm proposed in this paper, a robust algorithm due to Masuda and Yokoya and the RICP algorithm by Trucco et al. are evaluated.
We have experimentally determined the basin of convergence, robustness to noise and outliers, and computation time of these four ICP based algorithms
Mehrstufige Architekturen mit SQLJ und Enterprise JavaBeans Datenbankanbindung und Erstellung von Komponenten sind ein essentieller Bestandteil von Java-Anwendungen. Bei Java-Schnittstellen rückt der neue ANSI-Standard SQLJ für relationale und objektrelationale Datenbanken immer mehr in den Blickpunkt. Bei mehrstufigen Architekturen ist darüber hinaus die Komponentenbildung unabdingbar, und dabei entwickelt sich EJB zu einem De-facto-Standard. Nach einer grundlegenden Einführung geht es im zweiten Teil des Buches um Java-Schnittstellen zu Datenbanken: JDBC, ODMG und SQLJ8. Wie EJB-Komponenten erstellt werden und auf Java-Datenbanken zugreifen, wird im dritten Teil beschrieben.
We propose a reinforcement learning approach to heating control in home automation, that can acquire a set of rules enabling an agent to heat a room to the desired temperature at a defined time while conserving as much energy as possible. Experimental results are presented that show the feasibility of our method.
The growing size and complexity of software in embedded systems poses new challenges to the safety assessment of embedded control systems. In industrial practice, the control software is mostly treated as a black box during the system's safety analysis. The appropriate representation of the failure propagation of the software is a pressing need in order to increase the accuracy of safety analyses. However, it also increase the effort for creating and maintaining the safety analysis models (such as fault trees) significantly. In this work, we present a method to automatically generate Component Fault Trees from Continuous Function Charts. This method aims at generating the failure propagation model of the detailed software specification. Hence, control software can be included into safety analyses without additional manual effort required to construct the safety analysis models of the software. Moreover, safety analyses created during early system specification phases can be verified by comparing it with the automatically generated one in the detailed specification phased.
INSiDER: Incorporation of system and safety analysis models using a dedicated reference model
(2016)
In order to enable model-based, iterative design of safety-relevant systems, an efficient incorporation of safety and system engineering is a pressing need. Our approach interconnects system design and safety analysis models efficiently using a dedicated reference model. Since all information are available in a structured way, traceability between the model elements and consistency checks enable automated synchronization to guarantee that information within both kind of models are consistent during the development life-cycle.
Mobile-access edge clouds provide distributed com-pute capacities for low-latency applications. 5G technology willpave the way for such mobile deployment scenarios. In thispaper, we propose an edge cloud infrastructure that supportslow-latency video analysis connected with bandwidth reductionfor a moving group of persons. As example, we consider a mobilebody camera scenario that monitors the situation in a certain areaand transmits it to an operations center. Our discussion focuseson three aspects: mobility support, low-latency video processing,and bandwidth reduction. For this, we propose a mobile edgecloud infrastructure with a central cloud. In order to optimizevideo processing we optimize the edge cloud device assignmentof the cameras depending on their movement by reassigning itto another cloud device. This requires live migration of ongoingvideo analysis between edge devices. Finally we discuss the useof a mobile central cloud.
This paper describes using a mobile robot, equipped with some sonar sensors and an odometer, to test navigation through the use of a cognitive map. The robot explores an office environment, computes a cognitive map, which is a network of ASRs [36, 35], and attempts to find its way home.
Ten trials were conducted and the robot found its way home each time. From four random positions in two trials, the robot estimated the home position relative to its current position reasonably accurately.
Our robot does not solve the simultaneous localization and mapping problem and the map computed is fuzzy and inaccurate with much of the details missing.
In each homeward journey, it computes a new cognitive map of the same part of the environment, as seen from the perspective of the homeward journey. We show how the robot uses distance information from both maps to find its way home.
This paper shows how a mobile robot equipped with sonar sensors and an odometer is used to test ideas about cognitive mapping. The robot first explores an office environment and computes a "cognitive map" which is a network of ASRs [1]. The robot generates two networks, one for the outward journey and the other for the journey home.
It is shown that both networks are different. The two networks, however, are not merged to form a single network. Instead, the robot attempts to use distance information implicit in the shape of each ASR to find its way home. At random positions in the homeward journey, the robot calculates its orientation towards home. The robot's performances for both problems are evaluated and found to be surprisingly accurate.
In this paper, we present our experience in designing and teaching of our first robotics course for students at primary school level.
The course was carried out over a comparatively short period of time, namely 6 weeks, 2 hours per week. In contrast to many other projects, we use robots that researchers used to conduct their research and discuss problems faced by these researchers. Thus, this is not a behavioural study but a hands-on learning experience for the students.
The aim is to highlight the development of autonomous robots and artificial intelligence as well as to promote science and robotics in schools.
When animals (including humans) first explore a new environment, what they remember is fragmentary knowledge about the places visited. Yet, they have to use such fragmentary knowledge to find their way home. Humans naturally use more powerful heuristics while lower animals have shown to developa varietyof methodsthat tend to utilize two key pieces of information,namely distance and orientation information.
Their methods differ depending on how they sense their environment.
Could a mobile robot be used to investigate the nature of such a process, commonly referred to in the psychological literature as cognitive mapping? What might be computed in the initial explorations and how is the resulting “cognitive map” be used to return home?
In this paper, we presented a novel approach using a mobile robot to do cognitive mapping. Our robot computes a “cognitive map” and uses distance and orientation information to find its way home.
The process developed provides interesting insights into the nature of cognitive mapping and encourages us to use a mobile robot to do cognitive mapping in the future, as opposed to its popular use in robot mapping.
This paper presents a novel algorithm for computing absolute space representations (ASRs) in Yeap, W.K. and Jefferies, M. (1988) for mobile robots equipped with sonar sensors and an odometer. The robot is allowed to wander freely (i.e. without following any fixed path) along the corridors in an office environment from a given start point to an end point. It then wanders from the end point back to the start point. The resulting ASRs computed in both directions are shown
Generating a more detailed understanding of domestic electricity demand is a major topic for energy suppliers and householders in times of climate change.
Over the years there have been many studies on consumption feedback systems to inform householders, disaggregation algorithms for Non-Intrusive-Load-Monitoring (NILM), Real-Time-Pricing (RTP) to promote supply aware behavior through monetary incentives and appliance usage prediction algorithms. While these studies are vital steps towards energy awareness, one of the most fundamental challenges has not yet been tackled: Automated detection of start and stop of usage cycles of household appliances. We argue that most research efforts in this area will benefit from a reliable segmentation method to provide accurate usage information.
We propose a SVM-based segmentation method for home appliances such as dishwashers and washing machines. The method is evaluated using manually annotated electricity measurements of five different appliances recorded over two years in multiple households.
Shifting energy peak load is a subject that plays a huge role in the currently changing energy market, where renewable energy sources no longer produce the exact amount of energy demanded.
Matching demand to supply requires behavior Changes on the customerside, which can be achieved by incentives suchas Real-Time-Pricing (RTP).
Various studies show that such incentives cannot be utilized without a complexity reduction, e.g., by smart home automation systems that inform the customer about possible savings or automatically schedule appliances to off-peak load phases.
We propose a probabilistic appliance usage prediction based on historical energy data that can be used to identify the times of day where an appliance will be used and therefore make load shift recommendations that suite the customer’s usage profile. A huge issue is how to provide a valid performance evaluation for this particular problem.
We will argue why the commonly used accuracy metric is not suitable, and suggest to use other metrics like the area under the Receiver Operating Characteristic (ROC) curve, Matthews Correlation Coefficient (MCC) or F1-Score instead.
Real-world domestic electricity demand datasets are the key enabler for developing and evaluating machine learning algorithms that facilitate the analysis of demand attribution and usage behavior. Breaking down the electricity demand of domestic households is seen as the key technology for intelligent smart-grid management systems that seek an equilibrium of electricity supply and demand. For the purpose of comparable research, we publish DEDDIAG, a domestic electricity demand dataset of individual appliances in Germany. The dataset contains recordings of 15 homes over a period of up to 3.5 years, wherein total 50 appliances have been recorded at a frequency of 1 Hz. Recorded appliances are of significance for load-shifting purposes such as dishwashers, washing machines and refrigerators. One home also includes three-phase mains readings that can be used for disaggregation tasks. Additionally, DEDDIAG contains manual ground truth event annotations for 14 appliances, that provide precise start and stop timestamps. Such annotations have not been published for any long-term electricity dataset we are aware of.
Time series are series of values ordered by time. This kind of data can be found in many real world settings. Classifying time series is a difficult task and an active area of research. This paper investigates the use of transfer learning in Deep Neural Networks and a 2D representation of time series known as Recurrence Plots. In order to utilize the research done in the area of image classification, where Deep Neural Networks have achieved very good results, we use a Residual Neural Networks architecture known as ResNet. As preprocessing of time series is a major part of every time series classification pipeline, the method proposed simplifies this step and requires only few parameters. For the first time we propose a method for multi time series classification: Training a single network to classify all datasets in the archive with one network. We are among the first to evaluate the method on the latest 2018 release of the UCR archive, a well established time series classification benchmarking dataset.
Parameter free Non-intrusive Load Monitoring (NILM) algorithms are a major step toward real-world NILM scenarios. The identification of appliances is the key element in NILM. The task consists of identification of the appliance category and its current state. In this paper, we present a param- eter free appliance identification algorithm for NILM using a 2D representation of time series known as unthresholded Recurrence Plots (RP) for appliance category identification. One cycle of voltage and current (V-I trajectory) are transformed into a RP and classified using a Spacial Pyramid Pooling Convolutional Neural Network architecture. The performance of our approach is evaluated on the three public datasets COOLL, PLAID and WHITEDv1.1 and compared to previous publications. We show that compared to other approaches using our architecture no initial parameters have to be manually tuned for each specific dataset.
Dieser Artikel stellt ein Konzept f¨ur eine Online-Vorlesung im Rahmen der virtuellen Hochschule Bayern vor.
Zunächst werden kurz die fachlichen Inhalte erl¨autert, die es zu vermitteln gilt.
Nach dieser Einführung wird auf die Umsetzung der Inhalte in eine dem neuen Medium Internet angemessene Form eingegangen.
Dies betrifft vor allem die ausgewählten Techniken, die zur Wissensvermittlung eingesetzt werden.
Ein besonderer Schwerpunkt liegt auf der Möglichkeit durch Interaktivität das Wissen experimentell zu vertiefen und somit eine Verbindung zwischen dem theoretisch Gelernten und dem praktisch Erfahrenen herzustellen.
Weiterhin wird eine Vielfalt an Kommunikationsm¨oglichkeiten vorgestellt, da Kommunikation eine wichtige soziale Komponente im Lernprozess und h¨aufig entscheidend f¨ur den Erfolg ist.
In this paper we apply diagnostic analysis to gain a deeper understanding of the performance of the the keyword search system that we have developed for conversational telephone speech in the IARPA Babel program. We summarize the Babel task, its primary performance metric, “actual term weighted value” (ATWV), and our recognition and keyword search systems. Our analysis uses two new oracle ATWV measures, a bootstrap-based ATWV confidence interval, and includes a study of the underpinnings of the large ATWV gains due to system combination. This analysis quantifies the potential ATWV gains from improving the number of true hits and the overall quality of the detection scores in our system's posting lists. It also shows that system combination improves our systems' ATWV via a small increase in the number of true hits in the posting lists.
OBJECTIVES:
To generate a fast and robust 3-D visualization of the operation site during minimal invasive surgery.
METHODS:
Light fields are used to model and visualize the 3-D operation site during minimal invasive surgery. An endoscope positioning robot provides the position and orientation of the endoscope. The a priori un-known transformation from the endoscope plug to the endoscope tip (hand-eye transformation) can either be determined by a three-step algorithm, which includes measuring the endoscope length by hand or by using an automatic hand-eye calibration algorithm. Both methods are described in this paper and their respective computation times and accuracies are compared.
RESULTS:
Light fields were generated during real operations and in the laboratory. The comparison of the two methods to determine the unknown hand-eye transformation was done in the laboratory. The results which are being presented in this paper are: rendered images from the generated light fields, the calculated extrinsic camera parameters and their accuracies with respect to the applied hand-eye calibration method, and computation times.
CONCLUSION:
Using an endoscope positioning robot and knowing the hand-eye transformation, the fast and robust generation of light fields for minimal invasive surgery is possible.
The CALO Meeting Assistant (MA) provides for distributed meeting capture, annotation, automatic transcription and semantic analysis of multiparty meetings, and is part of the larger CALO personal assistant system. This paper presents the CALO-MA architecture and its speech recognition and understanding components, which include real-time and offline speech transcription, dialog act segmentation and tagging, topic identification and segmentation, question-answer pair identification, action item recognition, decision extraction, and summarization.
The CALO meeting assistant provides for distributed meeting capture, annotation, automatic transcription and semantic analysis of multiparty meetings, and is part of the larger CALO personal assistant system. This paper summarizes the CALO-MA architecture and its speech recognition and understanding components, which include real-time and offline speech transcription, dialog act segmentation and tagging, question-answer pair identification, action item recognition, decision extraction, and summarization.
In this paper, we describe a new Java framework for an easy and efficient way of developing new GUI based speech processing applications. Standard components are provided to display the speech signal, the power plot, and the spectrogram. Furthermore, a component to create a new transcription and to display and manipulate an existing transcription is provided, as well as a component to display and manually correct external pitch values. These Swing components can be easily embedded into own Java programs. They can be synchronized to display the same region of the speech file. The object-oriented design provides base classes for rapid development of own components.
Creating long-lived software systems requires a technol ogy to build systems with good maintainability. One of the core ideas of the Model Driven Architecture (MDA) is to ease the change of the run-time platform by raising the level of abstraction in which just the business aspects are modelled, and by separating business aspects from techni cal issues and implementation details. This article anal yses the MDA approach with respect to maintainability. We argue that MDA systems will become even harder to maintain because the maintainability depends on the sys tem's (development) environment. MDA, UML and other base technologies are still under development, therefore the tools will change considerably. While the MDA possibly eases the change of the run-time platform, we show that it is quite difficult to exchange a link in the development tool chain. Our argumentation is based on the general proper ties of software evolution and the dependency chains in the development and run-time environments. It is backed by experiences with MDA development as well as by analo gies to general maintenance experiences.
Abstraction is the most basic principle of software engineering. Abstractions are provided by models. Modeling and model transformation constitute the core of model-driven development. Models can be refined and finally be transformed into a technical implementation, i.e., a software system.
The aim of this book is to give an overview of the state of the art in model-driven software development. Achievements are considered from a conceptual point of view in the first part, while the second part describes technical advances and infrastructures. Finally, the third part summarizes experiences gained in actual projects employing model-driven development.
Beydeda, Book and Gruhn put together the results from leading researchers in this area, both from industry and academia. The result is a collection of papers which gives both researchers and graduate students a comprehensive overview of current research issues and industrial forefront practice, as promoted by OMG’s MDA initiative.
Es wird ein System aus der Augmented Reality (Erweiterte Realiäat, AR) vorgestellt, das zur Kamerakalibrierung einen farbigen Würfel verwendet, der außerdem als Platzhalter für ein virtuelles Objekt dient.
Zur Erkennung des Würfels in einer Szene werden Methoden aus der Farbbildverarbeitung verwendet, wie der Color Structure Code (CSC) und die Klassifikation der entstehenden Regionen nach ihrer Farbe. Zur Beschleunigung der Segmentierung wird ein hierarchisches Verfahren eingesetzt.
When animals (including humans) first explore a new environment, what they remember is fragmentary knowledge about the places visited. Yet, they have to use such fragmentary knowledge to find their way home.
Humans naturally use more powerful heuristics while lower animals have shown to develop a variety of methods that tend to utilize two key pieces of information, namely distance and orientation information.
Their methods differ depending on how they sense their environment. Could a mobile robot be used to investigate the nature of such a process, commonly referred to in the psychological literature as cognitive mapping? What might be computed in the initial explorations and how is the resulting “cognitive map” be used for localization?
In this paper, we present an approach using a mobile robot to generate a “cognitive map”, the main focus being on experiments conducted in large spaces that the robot cannot apprehend at once due to the very limited range of its sensors. The robot computes a “cognitive map” and uses distance and orientation information for localization.
We present an approach for indoor mapping and localisation using sparse range data, acquired by a mobile robot equipped with sonar sensors.
The chapter consists of two main parts. First, a split and merge based method for dividing a given metric map into distinct regions is presented, thus creating a topological map in a metric framework. Spatial information extracted from this map is then used for self-localisation on the return home journey.
The robot computes local confidence maps for two simple localisation strategies based on distance and relative orientation of regions. These local maps are then fused to produce overall confidence maps.
We present an approach for indoor mapping and localization with a mobile robot using sparse range data, without the need for solving the SLAM problem.
The paper consists of two main parts. First, a split and merge based method for dividing a given metric map into distinct regions is presented, thus creating a topological map in a metric framework.
Spatial information extracted from this map is then used for self-localization. The robot computes local confidence maps for two simple localization strategies based on distance and relative orientation of regions.
The local confidence maps are then fused using an approach adapted from computer vision to produce overall confidence maps. Experiments on data acquired by mobile robots equipped with sonar sensors are presented.
We present a novel split and merge based method for dividing a given metric map into distinct regions, thus effectively creating a topological map on top of a metric one. The initial metric map is obtained from range data that are converted to a geometric map consisting of linear approximations of the indoor environment.
The splitting is done using an objective function that computes the quality of a region, based on criteria such as the average region width (to distinguish big rooms from corridors) and overall direction (which accounts for sharp bends).
A regularization term is used in order to avoid the formation of very small regions, which may originate from missing or unreliable sensor data. Experiments based on data acquired by a mobile robot equipped with sonar sensors are presented, which demonstrate the capabilities of the proposed method.
We present an Approach for non linea roptimization of the parameters of an endoscopic camera mounted on a surgery robot. The goal is to generate a depth map for each image in order to enhance the quality of medical light fields.
The pose information provided by the robot is used as an initialization, where especially the orientation isi naccurate. Refinement of intrinsic and extrinsic camera parameters is performed by minimizing the back-projectionerror of 3-D points that are reconstructed by triangulation from image Feature stracked over an image sequence.
Optimization of the camera parameters results in an enhancement of Rendering Quality in two ways: More accurate parameters lead to better interpolation as well as to better depth maps for approximating the scenegeometry.
The paper presents an extended hand-eye calibration approach that, in contrast to the standard method, does not require a calibration pattern for determining camera position and orientation. Instead, a structure-from-motion algorithm is applied for obtaining the eye-data that is necessary for computing the unknown hand-eye transformation.
Different ways of extending the standard algorithm are presented, which mainly involves the estimation of a scale factor in addition to rotation and translation. The proposed methods are experimentally compared using data obtained from an optical tracking system that determines the pose of an endoscopic camera.
The approach is of special interest in our clinical setup, as the usage of an unsterile calibration pattern is difficult in a sterile environment.
The paper presents a new vectorquantization based Approach for selecting well-suited data for hand-eye calibration from a given sequence of hand and eye movements.
Data selection is essential if control of the movements used for calibration is not possible, especially when using continuously recorded data. The new algorithm is compared to another method for data selection as well as to the processing of subsequent movements.
Experimental results on real and synthetic data sets show the superior performance of the new approach with respect to calibration errors and computation time.
Real data has been obtained from an optical tracking system and a camera mounted on an endoscope, the goal being the reconstruction of medical lightfields.
This paper presents an approach for applying a dual quaternion hand–eye calibration algorithm on an endoscopic surgery robot. Special focus is on robustness, since the error of position and orientation data provided by the robot can be large depending on the movement actually executed.
Another inherent problem to all hand–eye calibration methods is that non–parallel rotation axes must be used; otherwise, the calibration will fail.
Thus we propose a method for increasing the numerical stability by selecting an optimal set of relative movements from the recorded sequence.
Experimental evaluation shows the error in the estimated transformation when using well–suited and ill–suited data. Additionally, we show how a RANSAC approach can be used for eliminating the erroneous robot data from the selected movements.
We describe an Augmented Reality system using the corners of a color cube for camera calibration. In the augmented image the cube is replaced by a computer generated virtual object.
The cube is localized in an image by the CSC color segmentation algorithm. The camera projection matrix is estimated with a linear method that is followed by a nonlinear refinement step.
Because of possible missclassifications of the segmented color regions and the minimum number of point correspondences used for calibration, the estimated pose of the cube may be very erroneous for some frames; therefore we perform outlier detection and treatment for rendering the virtual object in an acceptable manner.
This work presents a technique for computing dense disparity maps from a binocular stereo camera system. The methods are applied in an Augmented Reality setting for combining real and virtual worlds with proper occlusions. The proposed stereo correspondence technique is based oil area matching and facilitates an efficient strategy by using the concept of a three-dimensional similarity accumulator whereby occlusions are detected and object boundaries are extracted correctly. The main contribution of this paper is the way we fill the accumulator using absolute differences of images and computing a mean filter on these difference images. This. is. where the main advantages of the accumulator approach can be exploited, since all entries can be computed in parallel and thus extremely efficient. Additionally, we-perform an asymmetric correction step and a post-processing of the disparity maps that maintains object edges.
In this paper we address the problem of using quaternions in unconstrained nonlinear optimization of 3-D rotations. Quaternions representing rotations have four elements but only three degrees of freedom, since they must be of norm one.
This constraint has to be taken into account when applying e.g. the Levenberg-Marquardt algorithm, a method for unconstrained nonlinear optimization widely used in computer vision. We propose an easy to use method for achieving this.
Experiments using our parametrization in photo grammetric bündle -adjustment are presented at the end of the paper.
This paper presents new vector quantization based methods for selecting well-suited data for hand-eye calibration from a given sequence of hand and eye movements.
Data selection can improve the accuracy of classic hand-eye calibration, and make it possible in the first place in situations where the standard approach of manually selecting positions is inconvenient or even impossible, especially when using continuously recorded data.
A variety of methods is proposed, which differ from each other in the dimensionality of the vector quantization compared to the degrees of freedom of the rotation representation, and how the rotation angle is incorporated.
The performance of the proposed vector quantization based data selection methods is evaluated using data obtained from a manually moved optical tracking system (hand) and an endoscopic camera (eye).
The main focus of this work is the development of new methods for the self-calibration of a rigid stereo camera system. However, many of the algorithms introduced here have a wider impact, particularly in robot hand-eye calibration with all its different areas of application. Stereo self-calibration refers to the computation of the intrinsic and extrinsic parameters of a stereo rig using neither a priori knowledge on the movement of the rig nor on the geometry of the observed scene.
The stereo parameters obtained by self-calibration, namely rotation and translation from left to right camera, are used for computing depth maps for both images, which are applied for rendering correctly occluded virtual objects into a real scene (Augmented Reality).
The proposed methods were evaluated on real and synthetic data and compared to algorithms from the literature. In addition to a stereo rig, an optical tracking system with a camera mounted on an endoscope was calibrated without a calibration pattern using the proposed extended hand-eye calibration algorithm.
The self-calibration methods developed in this work have a number of features, which make them easily applicable in practice: They rely on temporal feature tracking only, as this monocular tracking in a continuous image sequence is much easier than left-to-right tracking when the camera parameters are still unknown.
Intrinsic and extrinsic camera parameters are computed during the self-calibration process, i.e., no calibration pattern is required. The proposed stereo self-calibration approach can also be used for extended hand-eye calibration, where the eye poses are obtained by structure-from-motion rather than from a calibration pattern.
An inherent problem to hand-eye calibration is that it requires at least two general movements of the cameras in order to compute the rigid transformation.
If the motion is not general enough, only a part of the parameters can be obtained, which would not be sufficient for computing depth maps. Therefore, a main part of this work discusses methods for data selection that increase the robustness of hand-eye calibration. Different new approaches are shown, the most successful ones being based on vector quantization.
The data selection algorithms developed in this work can not only be used for stereo self-calibration, but also for classic robot hand-eye calibration, and they are independent of the actually used hand-eye calibration algorithm.
Das Buch richtet sich an Studierende der Informatik oder verwandter Studiengänge und enthält Übungsaufgaben mit Lösungen aus Gebieten, die typischerweise in den ersten Semestern als Grundlagen behandelt werden. Ausgenommen ist der Bereich des Programmierens. Das Buch ergänzt den Grundkurs Informatik mit Übungen zu ausgewählten Kapiteln, ist aber auch in Kombination mit anderen Lehrbüchern verwendbar.
C/C++ GE-PACKT
(2007)