captured with an
It covers the entire
based image cloud
he ADS, i.e. nadir/
ie results of which
n section 4.1. RGB
nfo cloud based on
ed info cloud from
wn in Figure 5.
tol (Figure 5, bot-
g edges, with only
dow areas next to
rrectly covered by
hallenging for our
| general (cp., e.g.,
ls and below trees
a, consists of three
> stereo overlap is
Geosystems for the
Info cloud results
gure 6.
om RCD30 data in
(bottom).
Compared to the ADS result (Figure 5), the RCD30 TIN shown
in Figure 6, bottom, shows a higher noise level, on both the
paved ground and roof areas. This is due to the fact that there is
no redundant stereo coverage in this area, which would allow
for consistency checks, noise reduction and gap filling. Gaps
due to occlusions cause inclined rather than vertical TIN mesh-
es, visible at some building walls. However, such issues could
easily be addressed with a flight configuration that provides
more image overlap, e.g. 80%. Aside from that, the building
edge representation is correct and small details are very well
recognizable.
Figure 7: RGB colored info cloud of a DMC-II 140 stereo over-
lap in Aalen, Germany (top); TIN view of the Z/I Imaging buil-
ding (center); and enlarged info cloud of cars parked in front of
that building to illustrate the high point density (bottom).
21
5.3 Aalen (DMC-II 140)
A high-resolution block of the town of Aalen, Germany, was
used for SGM verification for the DMC-II. The image overlap
is approximately 60% along strip and 20% across strip, which
means most areas are covered by one stereo pair and, accor-
dingly, in a single info cloud only. An example result is shown
in Figure 7.
Based on the GSD of 5 cm, the SGM-derived info cloud fea-
tures more than 300 points/m? and, accordingly, contains a lot
of detail — such as street lamps (Figure 7, center) —, which is
especially visible when zooming all the way to the individual
point level for the cars parked in front of the Z/I Imaging buil-
ding (Figure 7, bottom). Similar to the RCD30 example dis-
cussed above, additional stereo coverage could be expected to
reduce noise and fill gaps, e.g. at the back of the van on the left.
Figure 8: Comparison of RGB (top) and FCIR (bottom) info
clouds in a forest area in Georgian Bay, Canada.
