Development of an Augmented Reality system using ARToolKit

Development of an Augmented Reality system using ARToolKit and user invisible markers
3. ARToolKit and the AR development process
ARToolKit is a software library written in the C programming language for creating AR
applications [Kat00] [Kat99]. It is distributed free for non-comercial or research
applications. It has been primarily developed by Dr. Hirokazu Kato of Osake
University, Japan, and is supported by the HIT Lab at the Univeristy of Washington and
the HIT Lab NZ at the University of Canterbury, New Zealand. Distributions for SGI
Irix, Linux, MacOS and Windows OS are available, and are distributed with complete
source code.
ARToolKit can be used for the easy development of a wide range of AR applications.
As discussed earlier, one of the key difficulties in developing AR applications is the
problem of tracking the user´s viewpoint. ARToolKit uses computer vision algorithms
to solve this problem by calculating the real camera position and orientation relative to
physical markers in real time. Also the other two elements of AR, registration and
calibration, are contained within ARToolKit. Registration is done by overlaying virtual
objects on the recognized physical markers; either created in VRML or OpenGL.
Calibration is provided by ARToolKit in the form of an easy camera calibration
algoritm.
For developing and running AR applications with ARToolKit the minumum hardware
requirements are a computer with a camera connected to it. Also can be used HMDs and
handheld displays. Both the use of video see-through and optical see-through HMDs are
supported. The computer should at least have a 500Mhz processor and a graphics
accelerator card. For the Windows platform USB cameras, FireWire cameras and
composite video cameras connected to a frame grabber can be used, as long as they
have DirectShow drivers.
3.1 Tracking and registration
Tracking is performed as a form of optical tracking; video frames are being grabbed and
searched for fiducial square markers. Such real world markers are pre-trained pattern
templates. The position and orientation of these markers is used for overlaying virtual
objects onto video frames. This gives the user a real time augmented view on the real
world. For being able to recognize the markers several steps are executed. First the live
video image is turned into a binary image based upon a preset lighting treshold value.
The resulting image is searched for squares regions. For every found square the pattern
inside it is matched against pattern templates. If a match occurs a marker has been
found. The next step is to calculate the position of the real camera relative to the
physical marker. These real world coordinates are used to set the position of the virtual
camera coordinates. This assures that the virtual objects drawn are precisely overlaid.
Figures 3.1 through 3.3 show the three steps involved in overlaying virtual objects.
Figure 3.1 Live video Figure 3.2 Binary image Figure 3.3 Virtual object
overlaid on live video
32