Images and Camera Calibration Data

Camera Perspective Projection Matrices

All five cameras are modeled by perspective projection model with radial distortion and fully calibrated in the same coordinate frame (a coordinate frame attached to the stereo rig). Radial distortion model is very simple: 
 [U]   [u - u0]                   [u_0]
 [V] = [v - v0] * (1 + k_1*r^2) + [v_0],  r^2 = (u - u_0)^2 + (v - v_0)^2
where [U,V] are image coordinates (in pixels) after radial distortion.

Each of the R,G,B channels of the three-chip color camera are calibrated separately.

black and white RGB color
Projection Matrix camera 1 camera 2 camera 3 camera 4 R G B
Distortion Coef. -1.06e-07 -1.06e-07 -8.77e-08 -8.37e-08 -9.62e-08 -9.54e-08 -9.02e-08

3D points are reconstructed in this "stereo-rig coordinate frame" and then transformed by the registration transformation (a rigid motion).

The Mapping between Coordinate Systems

There are three different coordinate frames:
  1. Stereo-rig local coordinate frame F1
  2. Room-segment local coordinate frame F2
  3. Room global coordinate frame F3
All cameras are fully calibrated in the stereo-rig frame. For each of the views within a room segment there is one rigid motion from the stereo-rig coordinate frame to the room-segment coordinate frame. See the scan-plan. For each segment there is a rigid mapping from the segment coordinate frame to the room global coordinate frame.

A point xi in coordinate frame Fi is transformed to point xi+1 in coordinate frame Fi+1 by a translation t followed by a rotation R:

xi+1 = R.xi + t.

The mapping is represented by a 3x4 matrix T = [R | t].

Mapping F1 -> F2 for a Given View
Scan aScan bScan cScan d Scan eScan fScan gScan h
Segment A 0, 1, 2, 3 1, 2 1, 2, 3 2, 3, 4 2, 3 1, 2 2, 3 3, 4
Segment B
Segment C 0, 1, 2, 3 1, 2 1, 2, 3 2, 3, 4 2, 3 1, 2 2, 3 3, 4
Segment D

Mapping F2 -> F3 for a Given Segment
Segment A Segment B Segment C Segment D
Go to: [Up]
Radim Sara
Last modified: Mon Mar 16 15:02:16 MET 1998