Control
H/V
[um] [m] [m] [m]
St. Gallen 39/47 3201 1105
11.1 0.60 0.60 0.68
Zug 36/47 2239 785
9.9 0.35 0.33 0.40
Table 2: Bundle adjustments without GPS
block adj | Control | Check | o,
RMS X, | RMS Y, | RMS Z, | p, Wr uL,
H/V H/V | [um]
[m] [m] [m] | [m] | [m] | [m]
St. Gallen | GPS | 4/11 30/47 | 11.1 | 0.59 0.57 0.13 | 0.93 | 0.81 | 0.96
Zug GPS | 4/16 36/47 | 11.4 | 0.57 0.44 0.18 | 0.72] 0.72 | 0.96
Table 3: Combined adjustments
which corresponds to 8.5 pm and 0.25 m in
planimetry and 0.5 m in height at a photo scale of 1:
27'000. Thus, compared to what is achivable the RMS
values in table 2 are slightly worse. Also, there were
slight differences of the results compared to the
adjustments with HATS (see above) which could be
attributed to different weighting of control points.
In the combined adjustment 12 resp. 5 station
coordinates observations from the St. Gallen resp. Zug
block were eliminated from the adjustment due to
gross errors. The RMS value of the GPS photo centres
are about 0.6 m in X, and Y,, and better than 0.2 m in
height. When using all control points as check points
the combined adjustment yields an empirical accuracy
(Uy) Which is by a factor 1.5-2 worse than the RMS
values from the reference adjustment.
These results show that the quality of the ground
control points is not very good. Specially, many of the
ground control points taken from the cadastre maps
were not of sufficient quality as well as the fact that
the cadastre was partially out of date. Here, correct
weighting of control points was very important for the
adjustments. This demonstrates clearly that it is
necessary to perform the GPS supported triangulation
using only signalized points, GPS photo centres and
additional height control at overlapping strip ends.
Thus, further subblocks which are to be processed of
block Switzerland must be defined in an area of at
least four available signalized control points. To test
the quality of height points in LK25 maps, 83 well
distributed points were measured in the block St.
Gallen, which were triangulated with HATS, and used
as height check points. As a result the average
difference of all points was 1.4 m, which is slightly
worse than the u, = 0.96 m from the comparison of
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
the control points used as check points, with the
maximum difference being 4.1 m.
3.4. Time required
The following time was required to process the in
table 4 summarized AT processing steps.
The total elapsed time which was required excluding
scanning, ground control preparation and data transfer
for triangulation of block St. Gallen was 51 h. This
corresponds to 28 min per image. Benefitting from the
experiences of the first triangulation block all 82
images of block Zug were processed in 27 h, which
corresponds to 20 min per image resp. 24 images per
day (8 h). Taking only the measurement time into
account, we were able to triangulate 32 images of
block Zug in one shift. In comparison the large
differences in the time used for blunder detection and
manual measurement of ground control as indicated in
table 4 can be attributed to the better knowledge of the
user interfaces and to the improved interface of
version 3.1.1.2 while processing the second block. But
the reduced elapsed time for block Zug compared to
block St. Gallen must also be partially attributed to
the use of a sparse tie point pattern, which causes less
interactive point measurements. Compared to
conventional triangulation on analytical plotters this is
only a slight speed-up, but, in general, there is still
potential for improvements in digital triangulation
using HATS. In table 4 all lines in italics indicates
processes, which could be automated or where the
elapsed time could be reduced significantly from our
point of view.
To ir
robus
follo:
sumn
(1)
any
possi
antici
APM
point
(2)
the
signif
afterv
hectai
whicl
(3).
shoul
(4) 4
APM
