IMD 4003: Motion Capture

For IMD 4003: Computer Animation, taught by Chris Joslin

Contents

1. 1. The System
   1. 1.1 Cameras
   2. 1.2 Lens
   3. 1.3 The Room
   4. 1.4 The Suit
   5. 1.5 Markers
2. 2. Vicon IQ
   1. 2.1 File Types
   2. 2.2 Calibration
3. 3. Step-By-Step Process

1. The System

1.1 Camera

• Strobe: used to illuminate the markers with a "visible" wavelength of 623nm.
• On-Board Processor: each camera has a processor that collects information from the sensor, processes it, sends the information to the camera network switch and finally to Tarsus. This information is compressed as 2D data containing only greyscale reflections of the data.
• Filter: each camera has a filter attached to it between the lens and the image sensor to filter everything but 623nm wavelengths.

1.2 Lens

The lens has a Focal Length of 12.5mm, a Horizontal (↔) FoV of 54.22° and a Vertical FoV (↑↓) of 42.0°.

• The area of light projected onto the image sensor is smaller than the area of the complete image sensor. This is why we don't have full frame projection.
• Depth of Field: the nearest an object can be without being blurry is 0.9m. The farthest it can be is over 10m.

1.3 The Room

The room must be large enough because we can only use a small subset where all of the camera's views converge. All of the reflective surfaces (retro-reflective, reflective and active light sources) must be either removed or covered.

1.4 The Suit

The purpose of the suit is to provide a connection between the markers and the person we're tracking. It keeps the markers on with Velcro so that we don't have to use glue or tape (though they are still sometimes used). It also helps reduce the reflection from the skin.

1.5 Markers

Markers are approximately spherical and have retro-reflective tape on them. There are 4 types of markers:

1. 18mm: these are flexible and used for bodies.
2. 14mm: these are solid and use for objects (especially those expecting impact or contact with other objects in the scene.
3. 4mm and
4. 2mm: these are used for facial and finger capturing (they offer more precision).

The 18mm (the ones we use in class) appear as 7-8px from the centre of the floor. They appear as 3px (the smallest you want to get a complete circle) at approximately 9m.

2. Vicon IQ

2.1 File Types

These are some common file types you will come across:

• .x2d: contains raw 2D image data compiled from all of the cameras.
• .vtt: (Vicon Threshold Template) contains the masking data for each camera.
• .cp: (Camera Parameters) contains information about where cameras are in the 3D space, the camera distortion, etc.
• .trial: contains reconstructed data in 2D for all marker positions.
• .enf: contains information about the directory and the trial files within the directory.
• .v: contains the animation data connected to a skeleton.
• .vsk: contains the skeleton for the subject being captured.

2.2 Calibration

The purpose of calibration is to:

• link the cameras to the 3D environment;
• link the cameras to each other in this environment in order to calculate the triangulation of the markers; and
• determine the distortion of the lenses.

Distortion: the further the marker is from the centre of the image, the more it is distorted. By using a "Distortion Map," the software can correct the distortion as much as possible.

Wand: the wand is set up with 3 markers which the cameras know about (so the ratio between the markers is more important than their actual physical distance). By recording the movement, the software can set the position in the 3D space. This is useful for calculating the Distortion Map too. We need about 6000 (±10%) frames with the 240mm wand.

3. Step-By-Step Process

There are five major steps:

1. Data Management
   • Create the Project
   • Create the Day
   • Create the Session
2. Setup
   • Open Vicon IQ
   • Connect to Tarsus
   • Set the Strobe to 100
   • Set the Threshold to 10-20
   • Set the Gain to 1
   • Set the Circle Quality to 30-40
3. Calibration
   • Wand Wave with 240mm wand
   • Set the Origin with the L-Frame
4. ROM
   • Put on the suit and the markers
   • Capture the ROM
   • Open the .vst Template
   • Reconstruct the markers
   • Label the subject
   • Process the trajectories
   • Calibrate the skeleton
5. Subject Capture
   • Reconstruct
   • Fill the gaps
   • Kinematic fit
   • Export the subject