away beyond a threshold, the bundle adjustment will 
be re-calculated to establish a new outer orientation. 
4.5.3 Quality Report As little as possible of the 
statistical information is exposed to the operator, but 
mainly used in the internal self-diagnosis. The 
information from the various computations are 
saved in log-files if needed for a deeper analysis. 
5. STABILITY TEST OF ANALOGUE 
HIGH SPEED CAMERAS 
In order to get a deeper knowledge of the geometrical 
properties of a high speed film camera during the 
exposure phase, a stability test of a camera has been 
performed. The study was designed out of the 
following conditions and questions: 
- Which parameters are stable during the 
exposure 
- Cameras without fiducial marks will be used 
- How should any instability be compensated 
and handled 
5.1 Equipment and Conditions 
Two set of cameras were used for the test. Only one of 
the cameras were possible to process further because 
of exposure problems with the other. A 3D test field 
was filmed from two camera stations to achieve a 
stereo coverage. The film was digitized with the 
TrackEye film scanner and measured. Approximately 
70 points were measured in each frame of which app. 
55 were common in both left and right image. 
Camera: LOCAM, 500 frames/second 
Optics: Smitar 10 mm 
Digitizing: 6.2 um/pixel (square) 
52 Evaluation of the Measured Film 
Three frames in the beginning and three frames in 
the end were measured by the TrackEye system. Of 
these six frames, four were selected for the stability 
test; frames number 1,4,90 and 91. 
The stability test was carried out as three separate 
parts: 
i A separate calibration on the data from each 
image frame 
ii A calibration on frame 1 
and an outer orientation using bundle 
adjustment on frame 1 followed by a 3D 
calculation of frames 1,4,90 and 91 with fixed 
outer orientation parameters. The mean 
deviation from the known points were 
compared. 
iii A calibration on frame 1 
and an outer orientation using bundle 
adjustment on each frame with the principal 
point as unknown. A 3D calculation of each 
frame with the orientation parameters and 
the principal point from its outer orientation 
computation. 
5.3 Separate calibration of each camera 
The camera were calibrated on a known test field 
with 55 points in one plane and 30 points in various 
positions separated from the plane. The purpose of 
the test was to find out which parameters were stable 
during a whole image sequence and between 
sequences. 
The calibration parameters were computed as a single 
point resection in space with the parameters as extra 
unknowns. The calibrated parameters were 
  
- principal distance [pixels] 
- principal point [pixels] 
- radial distortion coeff., separated in x and y 
- scale difference between x and y 
  
  
table 1 Calibration Parameters 
The result from the calibration shows the following 
characteristics: 
53.1 Principal Distance The principal distance 
seems to be stable during one sequence and fairly 
stable between to sequences. It is not known if the 
camera were accidentally moved or re-focused 
between the two camera positions. The difference is 
small enough to be regarded as noise. The deviations 
referres to the first frame. 
  
Principal Distance [pixels] 
Frame 1 4 90 91 
PD PD dev PD dev PD dev 
  
Pos1 1639.4 1639.8 +0.4 1639.2 -0.2 1638.4 -1.0 
  
Pos2 1635.0 1634.4 -0.6 1634.1 -0.9 1634.8 -0.2 
  
table 2 Principal Distance 
5.3.2 Principal point The principal point is, as 
expected, not stable between different frames. This is 
probably due to the digitizing process, but how large 
parts that are coming from the optical system is 
impossible to say. The original idea, to calibrate the 
principal point for each frame, seems to be necessary. 
The deviations referres to the first frame.