solved parameters are shown in Table 4. It can be seen that the
standard errors of certain parameters are very big compared to [parameter estimate std. error
the parameter values themselves. The reason for that can be that
the chosen distortion model is too complicated for the used px 6.184062e+02 |1.125537e+00
camera. Maybe just the parameters px, py, c, o and k; would py 4.775455e+02 |2.995073e--00
have been enough to model the camera. ec 1.40261 1e+03 | 5.532361e+00
parameter estimate std. error a 9.830809e-01 | 9.389204e-03
px 6.170304e+02 |2.184800e+00 P -3.767164e-04 | 1.795160e-03
Dy 4.835049e+02 | 1.885422e+00 kı 1.226342e-07 | 5.110449e-09
€ 1.406630e+03 | 1.044922e+00 ks 2.153417e-14 | 1.478256e-14
a 9.999628e-01 | 1.864703e-04 k; — |-8.917501e-20 | 1.436700e-20
B __|-5-007041e-05| 1.789253e-04 P; |-1.115324e-06 | 3.976178e-07
k; -3.677413e-14 | 2.514506e-14
k; -1.857348e-20| 2.765898e-20 Table 5. Results of the first image set.
Di 1.166692e-06 | 2.442701e-07
p? -7.634132e-07 | 2.308410e-07
Table 4. Results of the test field calibration.
The first captured image set consisted of three horizontally
rotated images with 50% overlap (see Figure 3). Totally 102
corresponding image points were measured manually. The
iterative calculation converged well and the results can be seen
in Table 5. As can be noticed the estimated values of the
parameters b,k,k;,p, and p; have again very large standard
errors compared to their values. And the rest of the estimated
values are very similar to those obtained from the test field
calibration.
Figure 3. First test image set. The distance to the building was
20-30 meters.
For the second test five images again with 50% overlap were
taken (see Figure 4) and 223 image points were measured on
them. In this case the camera was 2-3 meters away from the
object. The solved parameters with their standard errors are
shown in Table 6. Again the parameter values are quite close to
those obtained earlier except py, which has shifted over twenty
pixels. One reason for this shift can be the short distance to the
object, which might cause a different focusing. Or then the
small non-concentricity of the camera causes some bias to the
results on very short distances.
Figure 4. Second image set. The distance to the object was 2-3
meters.
—588-
arameter
estimate
std. Error
Dx
6.164060e+02
2.715294e+00
PY
4.518342e+02
2.550095e+00
1.407404e+03
4.720895e+00
1.005287e+00
4.040044e-03
1.413043e-02
3.109527e-03
1.255189e-07
7.340497e-09
-1.178755e-15
2.964080e-14
-4.838147e-20
3.551263e-20
2.032601e-07
2.583520e-07
8.416285e-07
2.580397e-07
Table 6. Results of the second image set.
In Fi
calcu
show
COOrc
but c
very
supp:
the rz
In th
came
that 1
the i1
possi
Some
devia
longe
devia
The
Becai
occlu
on th
statio
more
35|
30
25
20
Radial distortion:
REFEI
Browi
Photo.
Fryer,
Intern
Vienn
Hartel
Rotati
Olof |
Verlag
