![](bg6.png)
Figure 11: Mini Golf demo.
6.1 Performance
The interface was designed to be sufficiently
accurate to provide intuitive control and create a
convincing visual integration between real and
rendered objects. The accuracy of wand motion and
rotation has been tested in favourable lighting
conditions. Measurements were made at a range of
distances from the camera, with observations
repeated multiple times and compared to ground
truth data. Results, seen in Table 1, predictably show
uncertainty of the position estimates increasing with
distance from the camera. Linear movement
estimates are reasonably accurate, compensating
well for the effect of perspective on motion parallel
and perpendicular to the camera. However, a
substantial degree of inaccuracy is seen in x-axis
rotation estimates at a significant distance from the
camera.
Far (75cm) Mid (55cm) Near
(40cm)
10cm X
Translatio
n
10.03±0.09 10.02±0.06 10.03±0.0
6
10cm Z
Translatio
n
9.98±0.59 8.37±0.20 9.91±0.09
45° Z
Rotation
44.96±0.32 44.94±0.23 45.68±0.2
0
45° X
Rotation
45.28±4.61 39.76±2.76 45.39±1.0
6
In less favourable conditions, the presence of
bright light sources, shadows on the tracked features,
or noise due to low light can degrade performance.
However, a sufficient degree of control can be
achieved in most realistic indoor situations in which
the system was tested. As the tracking system is
based on colour, tracking problems are most
noticeable in situations where the colour of the
tracked features is present in substantial areas of the
background or user.
7 CONCLUSION
This paper has described methods of tracking a
known object in 3D in a single camera, through
properties of recognized features. This tracking is
used to create genuine 3D user interfaces that can be
used for direct interaction with 3D environments
integrating real and simulated objects. These
interfaces are suitable for use in a home
environment, with current computing hardware.
The current system is limited in terms of the
range of objects that can be tracked, requiring
markers to provide identifiable features for tracking
that incorporates rotation. However, the scope of the
fundamental system is broad enough that it could be
substantially extended with future development, and
further demonstrate the implementation and use of
original forms of human-computer interaction.
REFERENCES
Alter, T. D., 1992. 3D Pose from 3 Corresponding points
under weak-Perspective Projection. Technical Report
1378, MIT Artificial Intelligence Laboratory.
Chung, J., Kim, N., Kim, G.J., & Park, C.M., 2002. Real
Time Motion Tracking System For Interactive
Entertainment Applications. Proceedings of the 5th
International Conference on System Simulation and
Scientific Computing.
Haralick, R.M., Sternberg, S.R., & Zhuang, X., 1987.
Image analysis using mathematical morphology. IEEE
Transactions on Pattern Analysis and Machine
Intelligence, Vol. 9, No. 4, pp. 532-550.
Hartley, R., Zisserman, A., 2000. Multiple View
Geometry In Computer Vision, Cambridge University
Press.
Immersion Corporation, http://www.immersion.com
Kim, K., Ramakrishna, R.S., 1999. Vision-Based Eye-
Gaze Tracking for Human Computer Interface.
Proceedings of the IEEE SMC’99 Conference, No.2,
pp. 324-329.
Nielsen, F., Nock, R., 2004. Approximating smallest
enclosing disks. Proceedings of the 16th Canadian
Conference on Computational Geometry, pp. 124-127.
Rosenfeld, A., Pfaltz, J.L., October 1966. Sequential
operations in digital picture processing. Journal of the
ACM, Vol. 13, No. 4, pp. 471-494.
Smith, A.R., 1978. Color Gamut Transform Pairs.
Proceedings of SIGGRAPH 78, pp. 12-19.
Sony EyeToy, http://eyetoy.com
Yang, M-H., Ahuja, N., 1999. Recognizing hand gesture
using motion trajectories. Proceedings of the IEEE CS
Conference on Computer Vision and Pattern
Recognition, pp. 468-472.
Table 1: Mean values and standard deviation for estimates
of movement in 3D space.
VISAPP 2006 - MOTION, TRACKING AND STEREO VISION
394