Motion Capture - Computer Animation: Algorithms and Techniques - page 192

FIGURE 6.5
Complete marker set [ 6 ].
automatically or manually. This can be done manually, for example, by noting the frame in each
sequence when a particular heel strike occurs. However, with manually synchronized cameras the
images could be half a frame off from each other.
6.5 Fitting to the skeleton
Once the motion of the individual markers looks smooth and reasonable, the next step is to attach them
to the underlying skeletal structure that is to be controlled by the digitized motion. In a straightforward
approach, the position of each marker in each frame is used to absolutely position a specific joint of the
skeleton. As a first approximation to controlling the motion of the skeleton, this works fine. However,
on closer inspection, problems often exist. The problem with using the markers to directly specify posi-
tion is that, because of noise, smoothing, and general inaccuracies, distances between joints of the skel-
eton will not be precisely maintained over time. This change in bone length can be significant. Length
changes of 10-20 percent are common. In many cases, this is not acceptable. For example, this can
cause foot-sliding of the skeleton (also known as skating ). Inverse kinematics used to lock the foot
to the ground can counteract this skating.
One source of the problem is that the markers are located not at the joints of the performers, but
outside the joints at the surface. This means that the point being digitized is displaced from the joint
itself. While a constant distance is maintained on the performer, for example, between the knee joint
and the hip joint, it is not the case that a constant distance is maintained between a point to the side of
the knee joint and a point to the side of the hip joint.
The one obvious correction that can be made is logically relocating the digitized positions so that
they correspond more accurately to the positions of the joints. This can be done by using marker
information to calculate the joint position. The distance from the marker to the actual joint can be deter-
mined easily by visual inspection. However, the problem with locating the joint relative to a marker is