snowfox computer animation - Motion Capture
 

Scott Hamm
3D Modeler/Animator
Instructional Learning Consultant
hamms@umich.edu
Motion Capture

Student Intern Wolf displayed in real time on the Vicon motion capture system.

The Dancing Robot

The Kinesiology Department required a number of student groups to each collect several sets of motion capture data on their chosen subject.

This particular project involved recording motions for several different line dances common to weddings (or at least common to the weddings these students attended).

I personally supervised this project from camera calibration and suit-up, with a typical 44 marker configuration, through the motion test, data collection and skeleton mapping.

The data collection with this and the other student projects I supervised turned out very clean. This is the unedited output from my own favourite capture of that session:
small(~2.8MB) _______________ large(~11.2MB)

With the participating students' and professor's permission, I later mapped the data in Alias Motion Builder to a robot model I cobbled together one afternoon for this purpose. I'm still exploring post-processing clean-up, which with this data set is plainly needed for the feet.

Motion Capture Explained

At right, a School of Dance student suited up for motion capture.

Readying both the motion capture volume and the subject involve several steps in order to get the best possible data.

These steps include outfitting the subject, placing the reflective markers on the suit, and calibrating the volume to ensure the cameras capture as many markers as consistently as possible.

Once the system is ready to record, the subject goes through a range of motion test, somewhat like a series of yoga movements to demonstrate the extent to which the actor will move in the environment.

Once this is complete, the operator checks the initial data to ensure that all markers are consistently visible to the cameras. If all is well, the markers will be linked to form a skeleton, if a skeleton configuration is required for the final data set.

Then the actual capture session can begin. It is good practice to begin and end every session in a T pose, head up and arms straight out. This provides a constant reference for post production data cleanup. A session involves one short, constant animation, like a person throwing a ball or walking. Different animations are usually linked together later with 3D animation software.

Session data is mapped to the skeleton created from the range of motion test and if the collected data is clean and consistent, little further work within Vicon is necessary.

 
The Pitcher

Another of my groups captured a pitcher from the Athletics Department throwing a ball. Click on a link to see them at actual speed.

For this project, only the torso, head and one arm were recorded. The students bypassed mapping the data to a skeleton model in Vicon, and instead mapped the recorded points directly to a 3D human skeleton model.


back to top