| 381
sity Five Pulnix TM6CN cameras With specially designed ring lights were located on one wall of the environment chamber
t of (fig. 2). A network of four CCD cameras was found to be suitable to obtain the necessary precision but, since the
, it panels could deflect by significant amounts, a fifth central camera was included as a precaution Camera lenses were
Ist bundle adjustment procedure. Use of the laboratory calibration results a
essential. Without prior values high correlation between estimated para
procedure. Such correlation was particularly signifi
Parameters estimated by this method were included i
meters would be present in the self-calibration
cant for the central camera fitted with a 6mm lens. Camera
n all subsequent adjustments as calibration constants.
Camera Principal X Yp (mm) k1 k2 k3
Number Distance (mm) (mm)
Camera 1 6.223 0.185 | 0.097 | 3.5x1072 | -1.4x10 4 22x10 | 3.1x10-4
Camera 2 8.575 0.198 | -0.088 | 1.1x10(3 | 7.4x105 | -3.7410-0 4.2x104 | 2.1x104
Camera 3 8.541 20.067 | -0.124 1 1.6x103 | 1.1x100 | 5.1x10-7 -1.9x1075
Camera 4 8.607 0.121 -0.149 | 1.6x10°3 | -5.6x108 | 277x107 -2.1x1074 | 1.7x104
Camera 5 8.649 0.060 | -0.153 | 1.8x103 | 223x105 | 19107 -2.3x107^ | 32x104
av. std.dev 0.008 0.011 0.011 | 6.6x10° | 6.1x10-6 | 1.7x10-7 2.3x10-5 | 1.6x10-5
Table 1. Camera calibration parameters and their standard deviations.
e contains an image of the same panel from viewpoint 5. The fall off in intensity seen in
e typical for an unsupported wood panel after a rapid change in humidity.
Fig. 4. A typical panel image from viewpoint 5 Fig. 5. An unsupported warped panel from viewpoint 5
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995