1024(V) pixels with a pixel size of 16 micron 
and a focal length of about 20 mm. 
The calibration of the stereo vision system was 
done using a targeted test field. The stereo pairs 
were taken at different distances and view 
angles. Thirteen control points and four check 
points were used. 
3.1.1 Calibration with and without the 
Relative Orientation Constraints 
This test was conducted to analyze the relative 
orientation constraints. Eight image pairs were 
used for this purpose. The adjustment was 
calculated twice; with and without the relative 
orientation constraints. The relative orientation 
parameters were derived for every image pair 
from the exterior orientation parameters 
obtained by the bundle solution. Tab. 1 and Tab. 
2 show the calibration results with and without 
the relative orientation constraints. Without the 
constraints, the relative orientation parameters 
are different from one image pair to another and 
they are inaccurate. 
The base length between the stereo cameras 
obtained with the relative orientation constraints 
is exactly the same as the length measured by 
tape ( 1.827m + 1 mm). It is obvious that the 
calibration with the relative orientation 
constraints gives the stable parameters. The 
relative orientation constraints improved the 
results especially when the geometry of the 
blocks and the distribution of control points are 
not good. 
3.1.2 Calibration with and without 
Additional Parameters 
In this case, the calibration of the stereo vision 
system was done using the eight digital stereo 
pairs of our test field. The bundle triangulation 
with the relative orientation constraints was 
computed twice, with and without additional 
camera parameters. The principal point and the 
focal length of both cameras were always 
treated as unknowns. Seventeen points are used 
in the adjustment, where thirteen were control 
points and four were used as check points to 
compare the accuracy. Results are presented in 
Table 3. 
The additional parameters improved the 
accuracy by a factor of two. The radial 
distortion parameter k1 is the most significant 
parameter. 
3.1.3 Position Accuracy of the Stereo 
Vision System 
We used the computed orientation parameters in 
an intersection program to determine the object 
coordinates from the image coordinate 
measurements. This corresponds to the 
positioning of points with the stereo-vision 
system of the GPSVan, independent of object 
space control. Again, the coordinate of the 
targets of the test field were used for 
comparison. The results are displayed in Tab. 4 
showing the RMS errors positioned from 
different distances. 
It is fair to say that the positioning accuracy 
with two Kodak DCS cameras are in the 10 cm 
level for objects closer than 30 m in front of the 
van. 
3.2 Global Positioning Accuracy 
To test overall accuracy of our mobile mapping 
system, we measured four control points with 
a global accuracy of *1cm. To check the 
positioning accuracy, all control points were 
located by the mobile mapping system. Table 5 
shows the difference from points positioned by 
the GPSVan's stereo vision system and their 
true values. The distances of points to the 
GPSVan are shown. All points are in the State 
Plane coordinate system (Ohio south, zone code 
3402). 
For this test, the Turbo Rogues GPS receivers 
were installed on the GPSVan. The rotation 
angles of the GPSVan were taken from the 
combined adjustment of GPS and inertial system. 
The offset between the stereo vision system and 
positioning systems was calibrated using the 
method discussed in section[2.3]. 
  
Image Pair Bx [m] By [m] Bz [m] 
AQ Ao AK 
  
  
  
All 1.826 -0.008 -0.060 
-O: 
  
04297 0.05933 0.02015 
  
  
  
  
  
  
  
Tab. 1 Relative orientation parameters obtained by adding the 
relative orientation constraints in a bundle solution 
484 
| Image | 
200 
201 
202 
203 
204 
206 
208 
209 
  
  
In this p 
algorithmi 
positionin« 
one foot i 
the req 
applicatio