Winter Augmented Reality Meeting 2007
26th/27th February 2007, Graz, Austria
Advance Program
Day One
10:00 - 11:30 - Keynote by Mark Billinghurst (HitLAB NZ)
The previously invited keynote speaker Blair MacIntyre (Georgia Institute of Technology) had to cancel...
11:30 - 12:00 - Manfred Klopschitz (Graz University of Technology)
SLAV (SIMULTANEOUS LOCALIZATION AND VISUALIZATION)
Fiducial markers are a cheap and precise way for solving the localization and orientation problem for many Augmented Reality applications. One of the drawbacks is to scale the marker placement, and therefore the initial calibration, up in large scale environments. The vision of the SLAV project is a building that surveys itself and offers an updated 3D model of its geometry and tracking infrastructure to mobile users. A first step towards this goal is to automate the fiducial marker calibration for buildings. Markers are detected with a computer vision system and integrated in a sparse 3D point reconstruction. This reconstruction is made using visual odometry while the camera (potentially on a robot platform) is surveying the building.
12:00 - 12:15 - Break
12:15 - 12:45 - Joseph Newman & Alexander Bornik (Graz University of Technology/University of Cambridge)
Tracking for Distributed Mixed Reality Environments
A taxonomy of existing systems, libraries and frameworks used for developing Mixed Reality and Ubiquitous Computing environments, is applied to identify the areas that share the characteristics of these two fields. An architecture is proposed that can be used to support a wide-range of Mixed Reality applications and allow them to inter-operate in a distributed fashion, whilst simultaneously supporting straightforward porting of legacy applications.
12:45 - 13:15 - Thomas Pintaric (Vienna University of Technology)
Affordable Infrared-Optical Pose-Tracking
We present a new low-cost infrared-optical pose-tracking system for room-sized virtual environments, consisting of 4-8 shutter-synchronized 1394-cameras with an optical bandpass filter and infrared illuminator. All image-processing is done in software on an attached workstation. Preliminary results indicate low latency (20-40 ms), minimal jitter (RMS less than 0.05 mm/0.02 deg.), submillimeter location resolution and an absolute accuracy of ± 0.5 cm. Currently, ten independent 6-DOF targets can be tracked in real-time with an update rate of 60Hz.
13:15 - 13:45 - Gerhard Reitmayr (University of Cambridge)
Towards a complete localisation system for urban environments
I will discuss initialisation strategies for our model-based visual tracking system for urban environments from GPS. The large variance of GPS positioning in urban environments precludes a direct use as initialisation of a visual tracking system which requires a small variance due to the local search employed by the model-based approach. By using a search strategy informed by the GPS position, magnetic compass and the local model structure we can estimate a sufficiently accurate initialisation pose to allow the visual tracking to commence.
13:45 - 15:00 - Lunch (at Mensa)
15::00 - 15:45 - Demo of the IMS Tracking System (in VRLab at ICG)
15:45 - 16:15 - Gabriele Bleser (Fraunhofer IGD)
Design and development of marker-less camera tracking systems at Fraunhofer IGD
Augmented reality and camera tracking, as one of its key requirements, are research topics for several years now. The goal of this presentation is to give an overview over IGD's activities in the area of marker-less optical camera tracking. Applications and results are shown throughout the presentation to enhance the understanding.
The class of model-based approaches, which need a complete 3D model of the target scene as input, builds the starting point for the presentation. Two subcategories are identified here. The learning-based approaches shift most of the processing time into an off-line phase requiring preprocessing steps like the calibration of reference images or the generation of keypoint databases. This reduces the on-line processing time and makes the tracking quite stable. But the maintenance of the off-line model can be time-consuming. The on-line approaches gather all appearance-based information during the tracking. Different model-based systems relying on the registration of contours and points are presented.
Afterwards the point-based on-line system is expanded to partially known environments, where also 3D geometry is recovered on-line.
As recent literature shows, the trend goes from pure vision-based to hybrid tracking systems, where the camera as main sensor is aided by additional sensors and a sensor fusion core computes the optimal pose estimate from all available measurements. Results of a hybrid inertial-vision-based system are shown in the final part of the presentation.
16:15 - 16:30 - Break
16:30 - 16:45 - Peter Keitler (Technische Universität München)
A Graphical Planning and Analysis Tool for Ubiquitous Tracking Environments
The specification of tracking environments in an abstract notation called
spatial-relationship graphs (SRG) is a major issue in ubiquitous tracking.
When SRGs are interpreted at run-time by the tracking software, the
implementation of algorithms can be decoupled from the concrete application.
Furthermore, provided that error covariances for the individual sensors are
known, the resulting registration error can be derived by mathematical means
using error propagation.
In this talk, we will report on the current state of implementing a graphical configuration and analysis tool for ubitrack environments. In a first step,
we designed and implemented a graphical SRG editor. Currently, we are
developing a database interface for tracking environments. In the future, we
want to add functionality for supporting the user in planning and monitoring
tracking accuracy.
16:45 - 17:00 Florian Echtler (Technische Universität München)
Hand tracking for enhanced gesture recognition on interactive multi-touch surfaces
Recently, interactive surfaces with multi-touch sensors based on FTIR
(frustrated total internal reflection) have seen increased attention in
research. In this talk, a new method of gathering data about the users'
gestures beyond simple binary touch information will be presented.
An infrared light source is placed above the table and used to create
shadows of hands and arms. By tracking these shadows, new interaction
techniques can be developed where several consecutive and disjoint surface
contacts can be traced back to the same user. An example application is a
virtual whiteboard with persistent color assignments for each participant.
17:00 - 17:30 Tom Drummond (University of Cambridge)
Augmented Reality: The binding problem and asymmetries in the world of ubiquitous computing
This talk will present a high level overview of the context for Augmented Reality, what it can provide and what constraints exist on it. The binding problem will be considered in several contexts and the implications for AR and User Interface design will be considered. The implications of various asymmetries that exist in the problem space will also be discussed.
17:30 End
20:00 Precencia Meeting
Day Two
9:30 - 10:00 - Eduardo E. Veas (Graz University of Technology)
Ontology Assisted eXploration
This talk presents Ontology Assisted eXploration, a system providing a workflow for the creation of virtual representations of the real world within an AR application. It is widely known that AR has grown very rapidly in the presentation of information. It has also been noted that there is a lack of content creation AR applications. This work presents a novel method for creation of content for numerous AR applications.
Other aproaches to the creation of content for AR have focussed on the actual creation of geometry within an AR application. In late works in VR and AR (Vidente, and others), it has been emphasized that other ``semantic'' information about objects is required even for correct rendering. The proposed approach tries to reuse existing geometry and emphasizes the creation of objects with ``meaning''. Applications can then use the created content according to their own needs. As new applications are built using existing content, it is expected that the need for new ``meanings'' for existing objects will arise, the proposed system provides a way to extend the properties of already created objects and adapt them to new uses.
10:00 - 10:30 Michael Grabner (Graz University of Technology)
Tracking via Classification of Local Features
10:30 - 11:00 - Björn Schwerdtfeger (Technische Universität München)
An Evaluation of Mobile Information Presentation Schemes (for supra-adaptive Logistics Applications)
In supra-adaptive logistics systems, operative workers must be able to quickly and precisely retrieve objects from large storage areas with minimal training. We explored concepts toward providing workers with location-based wayfinding and picking information on mobile and spatial displays, varying both visualization schemes (1D: textual descriptions, 2D: map-based overviews and 3D: AR-based arrows) and display devices (head-attached, hand-held and spatial displays). In a comprehensive user study we evaluated our schemes against each other. The results of this study did not show the best solution for supporting supra-adaptive Logistic Applications, but have brought up another question: Is everybody able to work efficiently with augmented reality? Is there even a classification of how fast people can“learn” augmented reality?
11:00 - 11:15 - Break
11:15 -11:45 - Gerhard Schall (Graz University of Technology)
Precise Tracking for Geospatial Augmented Reality
Handheld Outdoor Augmented Reality (AR) requires a relative accurate position and orientation tracking to register the virtual information (e.g. virtual urban 3D model) with the physical buildings and objects. The availability of geospatial models for AR from real-world data appears to be a key enabling factor for the success of applications for handheld devices.
Typically the user's outdoor pose is given in a user reference system by GPS and inertial tracking. A major challenge is to get highly accurate tracking information as well as to fit the different reference systems of the base data or urban 3D models to a common absolute reference system also used by the tracking systems.
11:45 - 12:15 - Manuel Huber (Technische Universität München)
A Data Flow Architecture for Ubiquitous Tracking
As ubiquitous tracking setups are potentially distributed, dynamic and
highly heterogeneous they pose many different and interesting problems. At
the same time there are additional requirements on data acquired from
tracking and sensor systems in order to be useful for augmented or virtual
reality applications, such as timeliness, low latency and accuracy
constraints.
In this talk we present a data flow framework for processing tracking and
senor data which meets many of these goals while maintaining a high degree
of generality. The setup is modeled as a general directed graph with
processing components as nodes and with the data flowing along its edges.
A special emphasis of this framework is the generic design of components
which minimizes code duplication by allowing the same component to process
data in either push or pull style. This is especially useful in the context
of automatic data flow generation by pattern applications. Despite its
dynamic nature the framework maintains type safety of the component
connections and is easy to use and extend.
12:15 - 12:45 - Mathis Csisinko (Vienna University of Technology)
Towards a Universal Implementation of 3D User Interaction Techniques
This paper presents a versatile - write once, use everywhere - approach of standardizing the development of three-dimensional user interaction techniques. In order to achieve a platform and application independent implementation of 3D interaction techniques (ITs), we propose to implement the related techniques directly in the tracking middleware. Therefore a widely used tracking framework was extended by a Python binding to allow straight forward scripting of ITs. We cluster existing 3D ITs, into those which can be fully, partly or not implemented in the tracking middleware. A number of examples demonstrate how various interaction techniques can quickly and efficiently be implemented in the middleware and are therefore fully independent of the underlying application. We hint at how this approach can be used to decouple menu system control from the application with the final goal to help establishing standards for 3D interaction.
12:45 - 13:30 - Demos (at ICG)
13:30 - 14:45 - Lunch at Mensa
14:45 - 15:15 - Daniel Wagner (Graz University of Technology)
State of the art Handheld Augmented Reality
Although Augmented Reality is a natural extension to existing computer gaming areas, there is currently no system that fulfils the needs for deployment "in the wild". In this talk I present a system architecture for multiuser augmented reality applications on low cost, off-the-shelf handheld devices. The system we developed is highly scalable to many users, runs stable, has low per-client costs and exposes an easy-to-use user interface. We implemented several games that were tested in user studies outside controlled, laboratory environments.
15:15 - 15:30 - Christian Pirchheim & Manuela Waldner (Graz University of Technology)
Deskotheque
In this talk we will present ideas for a collaborative working environment for multiple users. For that purpose a room will be equipped with multiple projectors and workstations. All visible displays will be merged into a single large desktop. Users can bring their own devices (e.g. laptops) whose displays and input devices will be integrated into the system dynamically. Within the large desktop application windows can be moved freely using all available interaction devices. Deskotheque will be based on a modified X window system which allows to use any X-compatible applications.
15:30 - 15:45 - Break
15:45 - 16:15 - Alistair Beresford (University of Cambridge)
TIME for Better Transport
The UK transport network needs help: congestion alone costs the economy 12bn GBP per annum! In the Transport Information Monitoring Environment (TIME) project, we are exploring ways in which technology can help reduce the economic and environmental impact of transportation. Traditional intelligent transport applications are vertically integrated; in contrast, we are building an open, pervasive computer system to gather transport data from many different types of sensor, over wired and wireless networks, and to make it available via an open but controlled interface. This will make applications quicker and cheaper to build, and enable integration and inference from a multitude of sensor sources.
16:15 - 16:45 - Sven Havemann (Graz University of Technology)
The 3DKIOSK project: Towards 3D presentations as a standard in museums
In museums the great challenge is to present digitized cultural artifacts in an unintrusive and gentle fashion. Technology must not put itself in focus. Still, the difficult problem of mass-compatible ("grandma compliant") 3D navigation needs to be solved. The talks shows first results of this work in progress.
16:45 Finishing Words
20:00 Social Event at Glöcklbräu