Motion capture systems

**Figure 9.24:** With a *motion capture* (*MOCAP*) system, artificial features are placed around the body of a human actor. The motions are extracted and matched to a kinematic model. Each rigid body in the model has an associated geometric model that is rendered to produce the final animated character. (Picture from Wikipedia user Hipocrite.)
$\begin{figure}\centerline{\psfig{file=figs/mocap.ps,width=\columnwidth}}\end{figure}$

Tracking systems for attached bodies use kinematic constraints to improve their accuracy. The most common application is tracking the human body, for which the skeleton is well-understood in terms of links and joints [374]. Such motion capture systems have been an important technology for the movie industry as the motions of real actors are brought into a virtual world for animation. Figure 9.24 illustrates the operation. Features, of the same kind as introduced in Section 9.3, are placed over the body and are visible to cameras mounted around the capture studio. The same options exist for visibility, with the most common approach over the past decade being to use cameras with surrounding IR LEDs and placing retroreflective markers on the actor.

**Figure 9.25:** (a) The hand model used by Leap Motion tracking. (b) The tracked model superimposed in an image of the actual hands. (Figures by Leap Motion.)
$\begin{figure}\begin{center} \begin{tabular}{cc} \psfig{file=figs/leaphand.ps,wi... ...ion.ps,width=2.5truein} \\ (a) & (b) \\ \end{tabular}\end{center} \end{figure}$

To obtain a unique pose for each body part, it might seem that six features are needed (recall P6P from Section 9.3); however, many fewer are sufficient because of kinematic constraints. Additional features may nevertheless be used if the goal is to also capture skin motion as it moves along the skeleton. This is especially important for facial movement. Many new MOCAP technologies are currently under development. For example, a system developed by Noitom captures human body movement solely by placing IMUs on the body. Some systems capture motion by cameras alone, as in the case of Leap Motion (see Figure 9.25) for hand tracking, and systems by Microsoft and 8i for full-body tracking by extracting contours against a green screen. Solutions based on modern depth sensors may also become prevalent in the near future. One challenge is to make highly accurate and reliable systems for low cost and installation effort.