Refine
H-BRS Bibliography
- yes (58)
Departments, institutes and facilities
- Fachbereich Informatik (58) (remove)
Document Type
- Conference Object (30)
- Article (14)
- Report (6)
- Part of a Book (3)
- Book (monograph, edited volume) (2)
- Doctoral Thesis (1)
- Master's Thesis (1)
- Other (1)
Year of publication
- 2012 (58) (remove)
Keywords
- 3D-Scanner (2)
- ARRs (2)
- Bag of Features (2)
- FDI (2)
- Hybrid systems (2)
- classifier combination (2)
- clustering (2)
- feature extraction (2)
- machine learning (2)
- object categorization (2)
In service robotics, tasks without the involvement of objects are barely applicable, like in searching, fetching or delivering tasks. Service robots are supposed to capture efficiently object related information in real world scenes while for instance considering clutter and noise, and also being flexible and scalable to memorize a large set of objects. Besides object perception tasks like object recognition where the object’s identity is analyzed, object categorization is an important visual object perception cue that associates unknown object instances based on their e.g. appearance or shape to a corresponding category. We present a pipeline from the detection of object candidates in a domestic scene over the description to the final shape categorization of detected candidates. In order to detect object related information in cluttered domestic environments an object detection method is proposed that copes with multiple plane and object occurrences like in cluttered scenes with shelves. Further a surface reconstruction method based on Growing Neural Gas (GNG) in combination with a shape distribution-based descriptor is proposed to reflect shape characteristics of object candidates. Beneficial properties provided by the GNG such as smoothing and denoising effects support a stable description of the object candidates which also leads towards a more stable learning of categories. Based on the presented descriptor a dictionary approach combined with a supervised shape learner is presented to learn prediction models of shape categories.
Experimental results, of different shapes related to domestically appearing object shape categories such as cup, can, box, bottle, bowl, plate and ball, are shown. A classification accuracy of about 90% and a sequential execution time of lesser than two seconds for the categorization of an unknown object is achieved which proves the reasonableness of the proposed system design. Additional results are shown towards object tracking and false positive handling to enhance the robustness of the categorization. Also an initial approach towards incremental shape category learning is proposed that learns a new category based on the set of previously learned shape categories.
The ability of detecting people has become a crucial subtask, especially in robotic systems which aim an application in public or domestic environments. Robots already provide their services e.g. in real home improvement markets and guide people to a desired product. In such a scenario many robot internal tasks would benefit from the knowledge of knowing the number and positions of people in the vicinity. The navigation for example could treat them as dynamical moving objects and also predict their next motion directions in order to compute a much safer path. Or the robot could specifically approach customers and offer its services. This requires to detect a person or even a group of people in a reasonable range in front of the robot. Challenges of such a real-world task are e.g. changing lightning conditions, a dynamic environment and different people shapes. In this thesis a 3D people detection approach based on point cloud data provided by the Microsoft Kinect is implemented and integrated on mobile service robot. A Top-Down/Bottom-Up segmentation is applied to increase the systems flexibility and provided the capability to the detect people even if they are partially occluded. A feature set is proposed to detect people in various pose configurations and motions using a machine learning technique. The system can detect people up to a distance of 5 meters. The experimental evaluation compared different machine learning techniques and showed that standing people can be detected with a rate of 87.29% and sitting people with 74.94% using a Random Forest classifier. Certain objects caused several false detections. To elimante those a verification is proposed which further evaluates the persons shape in the 2D space. The detection component has been implemented as s sequential (frame rate of 10 Hz) and a parallel application (frame rate of 16 Hz). Finally, the component has been embedded into complete people search task which explorates the environment, find all people and approach each detected person.
This paper compares the memory allocation of two Java virtual machines, namely Oracle Java HotSpot VM 32-bit (OJVM) and Jamaica JamaicaVM (JJVM). The basic difference of the architectures in both machines is that the JamaicaVM uses fixed-size blocks for allocating objects on the heap. The basic difference of the architectures is that the JJVM uses fixed size block allocation on the heap. This means that objects have to be split into several connected blocks if they are bigger than the specified block-size. On the other hand, for small objects a full block must be allocated. The paper contains both theoretical and experimental analysis on the memory-overhead. The theoretical analysis is based on specifications of the two virtual machines. The experimental analysis is done with a modified JVMTI Agent together with the SPECjvm2008 Benchmark.
In a research project funded by the German Research Foundation, meteorologists, data publication experts, and computer scientists optimised the publication process of meteorological data and developed software that supports metadata review. The project group placed particular emphasis on scientific and technical quality assurance of primary data and metadata. At the end, the software automatically registers a Digital Object Identifier at DataCite. The software has been successfully integrated into the infrastructure of the World Data Center for Climate, but a key was to make the results applicable to data publication processes in other sciences as well.
In der vorliegenden Arbeit wurde ein Verfahren zur Analyse von Molekülen auf Grundlage ihrer molekularen Oberfläche und lokalen Werte für physiko-chemische und topografische Eigenschaften entwickelt. Der als Kernkomponente der Analyse entwickelte Fuzzy-Controller kombiniert molekulare Eigenschaften und selektiert die für Wechselwirkungen relevanten Merkmale auf der Oberfläche. Die Ergebnisse des Fuzzy-Controllers werden für die Berechnung von 3D-Deskriptoren und für die Visualisierung der ermittelten Domänen auf der Oberfläche herangezogen. Es werden zwei Arten von Deskriptoren berechnet. Deskriptoren, welche Flächeninhalte und Zugehörigkeiten zu den spezifizierten Bindungsmerkmalen der Domänen darstellen, und Deskriptoren, welche die räumliche Anordnung der Domänen zueinander beschreiben. Die vom Fuzzy-Controller überarbeitete Oberfläche wird im VRML-Format zur Visualisierung und weiteren Bearbeitung zur Verfügung gestellt. Die berechneten Deskriptoren werden zur Ähnlichkeitsanalyse von Liganden und zur Suche von komplementären Bereichen an der Bindungsstelle einesRezeptors eingesetzt. MTX in protonierter Form und DHF, die an das Enzym DHF-Reduktase binden, und die Inhibitoren Sildenafil, Tadalafil und Vardenafil des Enzyms PDE-5A wurden unter Ähnlichkeitsaspekten analysiert. Bei der Bestimmung von komplementären Bindungsmerkmalen wird ausgehend von den Bindungsmerkmalen eines Liganden nach komplementären Bereichen in der Bindungstasche des Rezeptors gesucht. Als Anwendungsbeispiele werden die Bindungsstelle des Enzyms DHF-Reduktase aus den Komplexen mit MTX und DHF und des Enzyms PDE-5A aus den Komplexen mit Sildenafil, Vardenafil und Tadalafil betrachtet. Insgesamt haben die Anwendungsbeispiele gezeigt, dass der vorgestellte Fuzzy-Controller Bindungsmerkmale auf der molekularen Oberfläche identifiziert unddie darauf basierenden, rotations- und translationsinvarianten Deskriptoren zur Ähnlichkeitsanalyse und zur Suche von komplementären Bereichen angewendet werden können.
A bond graph representation of switching devices known for a long time has been a modulated transformer with a modulus b(t)∈{0,1}∀t≥0 in conjunction with a resistor R:Ron accounting for the ON-resistance of a switch considered non-ideal. Besides other representations, this simple model has been used in bond graphs for simulation of the dynamic behaviour of hybrid systems. A previous article of the author has proposed to use the transformer–resistor pair in bond graphs for fault diagnosis in hybrid systems. Advantages are a unique bond graph for all system modes, the application of the unmodified standard Sequential Causality Assignment Procedure, fixed computational causalities and the derivation of analytical redundancy relations incorporating ‘Boolean’ transformer moduli so that they hold for all system modes. Switches temporarily connect and disconnect model parts. As a result, some independent storage elements may temporarily become dependent, so that the number of state variables is not time-invariant. This article addresses this problem in the context of modelling and simulation of fault scenarios in hybrid systems. In order to keep time-invariant preferred integral causality at storage ports, residual sinks previously introduced by the author are used. When two storage elements become dependent at a switching time instance ts, a residual sink is activated. It enforces that the outputs of two dependent storage elements become immediately equal by imposing the conjugate3 power variable of appropriate value on their inputs. The approach is illustrated by the bond graph modelling and simulation of some fault scenarios in a standard three-phase switched power inverter supplying power into an RL-load in a delta configuration. A well-developed approach to model-based fault detection and isolation is to evaluate the residual of analytical redundancy relations. In this article, analytical redundancy relation residuals have been computed numerically by coupling a bond graph of the faulty system to one of the non-faulty systems by means of residual sinks. The presented approach is not confined to power electronic systems but can be used for hybrid systems in other domains as well. In further work, the RL-load may be replaced by a bond graph model of an alternating current motor in order to study the effect of switch failures in the power inverter on to the dynamic behaviour of the motor.
For the case when the abstraction of instantaneous state transitions is adopted, this paper proposes to start fault detection and isolation in an engineering system from a single time-invariant causality bond graph representation of a hybrid model. To that end, the paper picks up on a long-known proposal to model switching devices by a transformer modulated by a Boolean variable and a resistor in fixed conductance causality accounting for its ON resistance. Bond graph representations of hybrid system models developed in this way have been used so far mainly for the purpose of simulation. The paper shows that they can well constitute an approach to the bond-graph-based quantitative fault detection and isolation of hybrid models. Advantages are that the standard sequential causality assignment procedure can be a used without modification. A single set of analytical redundancy relations valid for all physically feasible system modes can be (automatically) derived from the bond graph. Stiff model equations due to small values of the ON resistance in the switch model may be avoided by symbolic reformulation of equations and letting the ON resistance of some switches tend to zero, turning them into ideal switches.
First, for two examples considered in the literature, it is shown that the approach proposed in this paper can produce the same analytical redundancy relations as were obtained from a hybrid bond graph with controlled junctions and the use of a sequential causality assignment procedure especially for fault detection and isolation purpose. Moreover, the usefulness of the proposed approach is illustrated in two case studies by its application to standard switching circuits extensively used in power electronic systems and by simulation of some fault scenarios. The approach, however, is not confined to the fault detection and isolation of such systems. Analytically validated simulation results obtained by means of the program Scilab give confidence in the approach.
In the realm of service robots recovery from faults is indispensable to foster user acceptance. Here fault is to be understood not in the sense of robot internal, rather as interaction faults while situated in and interacting with an environment (aka ex-ternal faults). We reason along the most frequent failures in typical scenarios which we observed during real-world demonstrations and competitions using our Care-O-bot III 1 robot. They take place in an apartment-like environments which is known as closed world. We suggest four different -for now adhoc -fault categories caused by disturbances, imperfect per-ception, inadequate planning or chaining of action sequences. The fault are categorized and then mapped to a handful of partly known, partly extended fault handling techniques. Among them we applied qualitative reasoning, use of simu-lation as oracle, learning for planning (aka en-hancement of plan operators) or -in future -case-based reasoning. Having laid out this frame we mainly ask open questions related to the applicability of the pre-sented approach. Amongst them: how to find new categories, how to extend them, how to as-sure disjointness, how to identify old and label new faults on the fly.
Traffic simulations are typically concerned with modeling human behavior as closely as possible to create realistic results. In conventional traffic simulations used for road planning or traffic jam prediction only the overall behavior of an entire system is of interest. In virtual environments, like digital games, simulated traffic participants are merely a backdrop to the player’s experience and only need to be “sufficiently realistic”. Additionally, restricted computational resources, typical for virtual environment applications, usually limit the complexity of simulated behavior in this field. More importantly, two integral aspects of real-world traffic are not considered in current traffic simulations from both fields: misbehavior and risk taking of traffic participants. However, for certain applications like the FIVIS bicycle simulator, these aspects are essential.
Traffic simulations for virtual environments are concerned with the behavior of individual traffic participants. The complexity of behavior in these simulations is often rather simple to abide by the constraints of processing resources. In sophisticated traffic simulations, the behavior of individual traffic participants is also modeled, but the focus lies on the overall behavior of the entire system, e.g. to identify possible bottle necks of traffic flow [8].
Using virtual environment systems for road safety education requires a realistic simulation of road traffic. Current traffic simulations are either too restricted in their complexity of agent behavior or focus on aspects not important in virtual environments. More importantly, none of them are concerned with modeling misbehavior of traffic participants which is part of every-day traffic and should therefore not be neglected in this context. We present a concept for a traffic simulation that addresses the need for more realistic agent behavior with regard to road safety education. The two major components of this concept are a simulation of persistent agents which minimizes computational overhead and a model of cognitive processes of human drivers combined with psychological personality profiles to allow for individual behavior and misbehavior.
Approximate clone detection is the process of identifying similar process fragments in business process model collections. The tool presented in this paper can efficiently cluster approximate clones in large process model repositories. Once a repository is clustered, users can filter and browse the clusters using different filtering parameters. Our tool can also visualize clusters in the 2D space, allowing a better understanding of clusters and their member fragments. This demonstration will be useful for researchers and practitioners working on large process model repositories, where process standardization is a critical task for increasing the consistency and reducing the complexity of the repository.
This project investigated the viability of using the Microsoft Kinect in order to obtain reliable Red-Green-Blue-Depth (RGBD) information. This explored the usability of the Kinect in a variety of environments as well as its ability to detect different classes of materials and objects. This was facilitated through the implementation of Random Sample and Consensus (RANSAC) based algorithms and highly parallelized workflows in order to provide time sensitive results. We found that the Kinect provides detailed and reliable information in a time sensitive manner. Furthermore, the project results recommend usability and operational parameters for the use of the Kinect as a scientific research tool.
We present our approach to extend a Virtual Reality software framework towards the use for Augmented Reality applications. Although VR and AR applications have very similar requirements in terms of abstract components (like 6DOF input, stereoscopic output, simulation engines), the requirements in terms of hardware and software vary considerably. In this article we would like to share the experience gained from adapting our VR software framework for AR applications. We will address design issues for this task. The result is a VR/AR basic software that allows us to implement interactive applications without fixing their type (VR or AR) beforehand. Switching from VR to AR is a matter of changing the configuration file of the application. We also give an example of the use of the extended framework: Augmenting the magnetic field of bar magnets in physics classes. We describe the setup of the system and the real-time calculation of the magnetic field, using a GPU.