Erik Herrmann, Christoph Meißner, Uwe Kloos and Gabriela Tullius (Reutlingen University, Reutlingen, Germany) accepted Poster at 2011 Joint Virtual Reality Conference 20-21 September 2011, Nottingham UK EuroVR-EGVE
Abstract:
In order to increase the immersion in virtual reality applications, optical tracking of movements is a common interaction method. However, professional optical tracking systems are very expensive and often proprietary. Therefore we introduce an open source based solution for pupils and students with the intention to enhance their understanding and usage of 3D optical tracking systems.
1. Introduction
In 2005 Reutlingen University has begun to set up an immersive virtual environment system with the objective to have a hands-on VR setting for learning and teaching. However installing a professional system is expensive and therefore cannot be used as a demonstration object for schools and student laboratories. With the current work we present a low cost way to establish tracking systems for virtual environments for educational purposes.
2. Our Approach
We work just with two comparatively cheap webcams with modified filters. Without infrared filters and additional infrared bandpass filters the accuracy of the tracking system is increased. This also improves the results of simple thresholding for image segmentation and in the broader sense blob detection. The equivalent to Bouguet’s Camera Calibration Toolkit implemented in OpenCV [Wil11] is used to calibrate, undistort and rectify the camera images as described in [Bra08]. Since we want to track infrared markers we need to track blobs of light in the images and find their correspondences. We use methods of the OpenCV Library for image segmentation and blob detection. For blob tracking and correspondence finding we implemented our own algorithms based on epipolar geometry. For better maintainability and understanding of our tracking system by students, our system is implemented in C# rather than in C++. Therefore we use EmguCV [Em11] as wrapper for OpenCV. Additionally the system uses VRPN [Vrp11] as interface so the tracking data can be used by client applications.
3. Conclusions
So far our work has shown the feasibility of the approach. It shows that the understanding of computer vision by students can be improved in contrast to those that never had any practical experience working with such a system. Before extending the system the next step is to evaluate the system with regard to its accuracy, reliability, and later to its performance. Another point of future work will be advanced research into ways to automate the calibration process.
References
[Bra08] Bradski, Gary; Kaehler, Adrian; Learning OpenCV: computer vision with the OpenCV library; O’Reilly; 2008
[Em11] www.emgu.com (last access: 09.06.2011)
[Vrp11] www.cs.unc.edu/Research/vrpn/ (last access: 09.06.2011)
[Wil11] opencv.willowgarage.com (last access: 09.06.2011)
Source Code ansehen