i
i
i
i
i
i
i
i
4.3. Technology for Motion Tracking and Capture 79
be sent to the computer, it may only be possible to sample positions and
orientations 10–20 times a second. This may not be fast enough to track
rapid movements accurately enough.
4.3.4 Optical Tracking
One method of optical tracking works by recognizing highly visible markers
placed on the objects to be tracked. If three or more cameras are used, the
marker’s 3D position can be obtained by simple triangulation and the known
properties of the cameras. However, to achieve greater accuracy and more
robust position determination, techniques such as those discussed in Chap-
ter 8 will need to be employed. Optical tracking of point sources cannot
provide orientation information, and it suffers badly from the light sources
being obscured by other parts of the scene. Overcoming such difficulties
might require several cameras. To obtain orientation information is more dif-
ficult. Some methods that have been tried include: putting several differently
shaped markers at each tracked point, or arranging for a recognizable pattern
or lights to be generated around the tracked point.
Many novel optical tracking ideas have been proposed. For example, sen-
sors on the tracked point may be excited when a laser beam which is contin-
ually scanning the scene strikes them (like the electron beam on a TV or a
bar-code reader in the supermarket). Laser ranging can give the distance to
the sensor. Three scanners will determine a position either by triangulation or
simply distance measures. Three detectors per sensed point will allow orienta-
tion information to be obtained. Because the speed of light is so much faster
than the speed of sound in air, an optical tracking system does not suffer any
of the acoustic delay.
As we shall see in Chapter 8, the science of computer vision and pattern
recognition offers a reliable way of tracking visually recognizable markers in
video images, which can be acquired from a network of cameras overlooking
the working volume. This idea comes into its own when we wish to add
virtual objects to the real-world view so t hat they appear to be part of the
scene, as happens in augmented reality. This form of optical tracking can
be used in practice by extending the open source augmented reality toolkit,
the ARToolKit [5]. For example, the markers that are attached to the objects
shown in the three camera views of Figure 4.21 can be tracked and located
very accurately if the locations of the cameras are known and they have been
calibrated. Camera calibration is discussed in Section 8.2.2, and a tracking
project is described in Chapter 18.