Information Technology Reference
In-Depth Information
degrees-of-freedom information from a 2D camera image. This principle is widely
used in augmented reality and tangible interaction systems. Several toolkits offer
out-of-the-box support for application developers. ARTag [28] supports up to 1024
barcode markers which have been optimized for fast and reliable detection. AR-
ToolkitPlus [162] is an improved version of the original ARToolkit
4
that is inspired
by ARTag's approach. It offers binary markers that can be more robustly detected
than the pattern-based markers of the original ARToolkit.
Topological region adjacency
is leveraged by a third class of fiducials. The basic
idea of this approach is, instead of encoding a sequence of bits as a sequence of black
and white points, to encode a hierarchical graph. The original bit sequence is de-
coded by identifying and then traversing this graph. The advantage of this approach
is that it is fast and very robust against false positive detection [104]. However, only
a small number of identifiers can be encoded. reacTIVision [53] is an open-source
toolkit that uses this technique. Figure 2.2 (e) shows a reacTIVision fiducial. reac-
TIVision is used in many projects that require tracking the location and orientation
of tangibles on interactive tabletops. While reacTIVision tracks the 2D position and
orientation of objects on a flat surface, it cannot provide full 6 degrees-of-freedom
information. Recent research showed how to obtain 6 degrees-of-freedom informa-
tion using topological region adjacency markers [104].
The main advantage of using fiducial markers for identifying and tracking objects
is that this is a relatively inexpensive tracking solution. However, this approach re-
quires that the fiducial is in the line-of-sight of the camera. This requirement can
severely restrict natural interactions, for instance, when objects are piled. Moreover,
the fiducial interferes with the artwork of the object.
Content-embedded fiducials
are visually less obtrusive. DataGlyphs [52] (Fig. 2.2
(e)) encodes binary data with a pattern of forward and backward slashes. They are
flexible in size, shape and color. This makes it possible to emulate the look of a
grayscale or even of a color image by an appropriate pattern of small dashes, simi-
lar to how offset printing emulates images by small raster dots. At 600 dpi printing
resolution, DataGlyphs can encode up to 1,000 bytes of data per square inch. An-
other technology, Anoto digital pen and paper [4], performs tracking by decoding a
pattern of tiny points that is printed onto paper documents and hardly visible to the
human eye. The Anoto approach is presented in more detail in Section 2.1.4 below.
Some applications require real-time location tracking in three dimensions, even
in cases when objects are moving very fast. For instance this is important in aug-
mented reality applications that overlay physical objects with projected digital con-
tents. Optical motion capture systems (e.g. Vicon
5
, OptiTrack
6
) use several high-
speed infrared cameras that observe a scene from different perspectives. Several
small retro-reflective dots are attached to the object which is to be captured. These
dots appear as white blobs in the camera images. If a dot is seen by at least two
4
http://www.hitl.washington.edu/artoolkit/
5
http://www.vicon.com
6
http://www.naturalpoint.com/optitrack
