International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
- Full: the complete, comprehensive block with 8 flight
lines (Figure 2) and all 12 GCPs.
- 2 GCP, I: a single bi-directional flight line with 9 images.
(total of 18 images), and a double GCP (2 XYZ-GCPs
located close to each other) (Figure 1, I-block)
- 0 GCP, I. a single bi-directional flight line with 9 images,
(total of 18 images), and no GCPs (Figure 1, I-block)
The calibration task consisted of the calibration of four
GPS/IMU/optics-combinations; NLS is operating two aircrafts
(OH-ACN, OH-CGW), both having RC20 cameras with ex-
changeable wide-angle (153 mm) and normal-angle (214 mm)
optics, and Applanix POSAV™ 510 systems (see Table 1).
NLS performed all the preprocessing of the data (GPS/IMU-
processing, photographic processing, scanning and image
measurements), and delivered the observation data to the FGI
for the further analysis.
Complete analysis of the direct exterior orientation data was
reported by Honkavaara et al. (2003). The results were in
accordance with the vendor's specifications (5.6 mgon in o and
©, and 8.9 mgon in K), excluding minor exceeding in two
blocks. Evaluation of the direct position observations was dis-
turbed by the poor quality of the reference values and the appa-
rent systematic errors (interior orientation, datum).
3.2 Methods
The investigation was made by evaluating the calibration
results of 10 test blocks (scales 1:8000 and 1:16000) using
three different block-structures (see Chapter 3.1). Calibrations
were performed using bundle block adjustment software
FGIAT of FGI. The mathematical models of FGIAT were
described in Chapter 2.1. FGIAT treats the image deformation
parameters as weighted observations and GPS/IMU-parameters
as free unknowns. The object coordinates were in a local
tangential coordinate system.
The qualities of the parameters were evaluated after
calibration. The significance of a certain parameter was evalua-
ted by comparing the parameter value to its standard deviation;
if the value was at least 2 times larger than the standard devia-
tion, the parameter was considered as significant. Also the
correlations of the calibration parameters were evaluated.
The DG accuracy was evaluated in the image space by calcula-
ting the image coordinates of the checkpoints using the calibra-
tion parameters and direct orientation observations and compa-
ring the calculated values to the measured values. GCPs were
used as checkpoints. The limitation of this approach was espe-
cially with the full blocks that the calibration and the quality
analysis were made with the same data. However, the analysis
gave understanding about the performance of various cases.
4. RESULTS
4.1 Determinability of parameters
Correlation of the radial distortion parameters (k1, k2, k3) was
remarkable, as expected. When all the radial distortion para-
meters were used, their correlations with interior orientation
parameters become significant. If only k1 was used, the corre-
ARS p
[ Comprehensive | [Cross] [1]
Al 4
La
Pi Ji.
Ex. »p
vv wv
Figure 1. Comprehensive, cross and l-block structures.
Sjókulla calibration field, E
scales 1:8000 and 1:16000
Figure 2. 1:8 000 and 1:16 000 scale calibration field of FGI.
Size: 4 km x 5 km, 12 accurate GCPs.
Date Blo- Optics
2002 ck Plane: OH-CGW Plane: OH-ACN
13026, 7163, 13153 7183,
153.300 214.066 153.030 214.108
244 2119 1:16 000
2120 1:8 000
25.4 2121 1:8 000
2122 1:4 000
26.4 2124 1:8 000
2. 2128 1:8 000
4 S. 2129 1:16 000
14.7. 2134 1:16 000
I5 7. 2135 1:16 000
3.9. 2137 1:16 000
4.9. 2136 1:16 000
Table 1. Calibration flights of NLS. The block structure and
the GCP distribution are shown in Figure 2.
lation was low. Correlations were high also between principal
point and tangential distortion parameters (x0 & pl, yO & p2).
An example of the correlations between the physical
parameters and boresight parameters is shown in Table 2 for
two different block structures. Tangential parameters had quite
high correlation with the boresight parameters (pl & d« and p2
& do). When tangential parameters were used, the correlations
between boresight and principal point parameters got high (yO
& do, x0 & de). When tangential parameters were not used
correlation of these parameters was low.
The block structure had some effect on the determinability of
the parameters. The affinity parameter (B1) could not be accu-
rately determined with the I-block. In I-block appeared also
high correlations between some Ebner's parameters and other
parameters. Comprehensive and cross-shaped blocks did not
have these problems.
The additional parameters were mostly significant; especially c
(optics 7183), yO (all optics), radial distortion (typically all the
terms), tangential distortion (typically both terms) and Ebner's
parameters (in general 1-3 insignificant parameters). The block
adjustment results were consistent when Ebner's parameters
were used. If the full set of physical parameters were included
in the adjustment, high correlations appeared with many para-
ci
ID
c x —d
m
ra
st
of
de
al
SI
ol
p:
a
de
