7 of or-
ey has
st cam-
its was
st site.
Irvey a
d. The
uation
! profile
ntifiable
ccurred
ammet-
> North-
s really
are the
the two
for this
f the 3-
ot easy
nt.
her with
ermined
by the Chair of Photogrammetry and Remote Sens-
ing, Technical University of Munich. The orientation
is estimated by a bundle solution for three-line im-
ages, in which a large number of tie points together
with 56 of ground control points is introduced. The
first results (Kornus et al., 1995) have been summa-
rized above. In the meantime a certain improvement
was achieved by this group so that for the recent ori-
entation an accuracy level of 6 to 8 m in all three
coordinates is expected.
2. DTM GENERATION
For DTM generation with MOMS stereo images a fea-
ture based procedure was proposed by Hahn and
Schneider (1991). In the meantime a matured soft-
ware module is developed which is highly efficient in
processing three-line images (Schneider and Hahn,
1995). The trinocular stereo solution consists of the
following steps: (1) Image pyramids are generated
for all channels and matching is started on a coarse
resolution level of the pyramid. Processing then pro-
ceeds from coarse to fine with (2) the extraction of
interest points and edge points; (3) the matching of
corresponding features of all three channels using
given orientation; (4) the determination of 3-D point
coordinates; and (5) surface reconstruction by finite
element modelling.
DTM generation is performed with scene 17 of the
Australia orbit 75B. An area of about 14.5 km x 14.5
km, in which the DTM evaluation profile can be seen,
is shown in figure 4.
fence
fence
jou 4: Ortho-image of an area of 14.5 km x 14.5
m
269
In this figure an ortho-image is displayed that is calcu-
lated from the forward looking channel. The markings
"fence" indicate a very long straight line, presumably
of a grazing fence. Along this fence there is a track on
which the 3-D profile is recorded. This track is visible
in the image (see figure 5).
Figure 5: Detailed view: the track along the fence can
be easily recognized.
A perspective view of the derived DTM is shown in
figure 6. In this case the ortho-image is overlayed to
enhance the visual impression of the terrain surface.
Because of the very flat terrain the height component
is amplified by a factor of 20. The grid width of this
DTM is 240 m.
3. INVESTIGATION OF THE DTM ACCURACY
So far the experimental investigations of the geomet-
ric accuracy in MOMS DTM reconstruction have been
very limited due to the lack of sufficient ground truth
data. The test field for evaluating the metric perfor-
mance of the stereo module (figure 2) was quite use-
ful for the analysis of sensor models and line-image
orientation but of very limited evidence in assessing
DTM accuracy.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996