International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
football field in Zwieselstein (Ötztal/Austria) was surveyed and
used as calibration area.
3.3 Investigation of the Vernagtferner data set
In August 2002 the Vernagtferner, one of the glaciers included
in the OMEGA project and monitored permanently by the
Commission of Glaciology of the Bavarian Academy of
Science, was covered by laser scanning. The final data set was
delivered several months after the flight date due to geo-
referencing problems.
The laser points have been transformed from UTM (WGS84)
into the Austrian Gauss-Krüger system, since the existing
control points and photogrammetric data are available in this
system. For accuracy assessments eight test areas have been
determined with different material properties (ice, snow, rock,
debris) and terrain inclination. Within these test areas check
points have been measured with differential GPS. The results
are given in table 3.
Test area Number of RMS error Offset
check offset corrected
points [m] [m]
Rock 1 70 i023 +0.12
Snow 59 + 0.08 +0.03
Debris 171 +0.14 -0.13
Ice 510 + 0.29 -0.21
Profile (ice) 40 £0.11 -0.19
Moraine 78 x 0.16 +0.80
Rock 2 37 + 0.25 +0.54
Rock blocks 128 £0.26 +0.78
Table 3. Results of the accuracy check by GPS check points
As common in high mountain surveying, triangulation points
are signalised by stone pyramids and usually are used as ground
control points in photogrammetry. It was possible to locate
some of these points within the laser data. The positions of the
top point of the pyramid formed by the laser data mainly
correspond properly to the position of the bench mark given by
ground coordinates. For checking the height value, the
difference to the “nearest neighbour” of the surrounding laser
points was computed at 7 bench marks. Table 4 shows that the
heights of the laser points obviously are lower than the
benchmark heights because the corresponding laser points
normally don’t match the top of the pyramid. Therefore instead
of stone pyramids other kind of control information like
geometrically plane objects (e.g. roofs) should be used (see also
Kraus, Pfeifer, 1998). However, such objects hardly can be
found in high mountain areas.
Bench mark Height difference (m)
STM 1 -0.70
STM 2 -0.74
STM 3 -0.71
STM 4 -0.75
STM 6 -0.40
Gabel (Hut) -0.70
STM 7 -0.76
Table 4. Height differences at bench marks (stone pyramids)
A further indication on the quality of laser data is given by
checking the consistency of overlapping strips. For this reason a
small test area within overlapping strips was investigated. After
756
generating a Im DEM contours with an interval of 0.5 m were
derived. In figure 3 the contours are superimposed with the
laser data from adjacent strips. It is obvious that the waves are
caused by the height differences of the adjacent strips. The
height difference can be estimated by graphical interpretation to
NE Lo WAAL)
“ME
AL SNR
UA | : Ms (NME C 1
4 X J qe - à
a
© UA
A
Xe
Ne
SD
» 2 t T M N ui ; 4 7
A ANN NN + b EH
Jas ANN AL
A
NE dc N
Figure 3. Contours showing “wave effect”
4. COMPARISON OF IMAGE MATCHING AND
LASER SCANNING
Within the OMEGA project an image flight over the
Vernagtferner glacier has been carried out in August 2003. The
colour images with a mean image scale of 1: 16.000, taken by a
standard RMK TOP camera (focal length 154 mm) have been
used for semi-automatic DEM generation (s. Chapter 2). The
exterior orientation was reconstructed using some of the bench
marks, also used for checking the laser data in order to ensure
. for consistent geo-referencing. This way the laser DEM and the
DEM from aerial images can be compared, taking into account,
that there is a time difference of one year.
4.1 Test area glacier tongue
A test area was defined, including the glacier tongue and the
forefront of the glacier. The captured laser data show
considerably varying point density. Sparse reflectance can be
recognized especially on the wet ice surface and on the streams
(see figure 4 and 5). A 2.5 m spaced DEM was generated both
from the laser data and the aerial images. The shaded relief
models (Figure 6 and 7) show more details in the laser DEM
than in the DEM from image matching.
For a detailed inspection height differences at the DEM grid
points have been calculated and colour coded (Figure 8). It is
obvious that there was a considerable height decrease at the
glacier tongue (up to 5 m) and at the upper right part of the test
area caused by melting debris covered ice. For accuracy
consideration only the non ice covered area can be investigated.
The colour coded height differences in this area show a constant
shift of about 0.5 m.
Calculated offsets between matched DEM, laser DEM and GPS
check points are presented in table 5. Offsets may be caused
either by an error of the exterior orientation of the aerial images
or by a systematic error of the laser points. It seems that laser
scanning is the more accurate method. Exact statements,
however, needs to evaluate the local behaviour of the different
height systems, i.e. geometrical heights for GPS measurements
and orthometric heights for photogrammetric ground control.
The difference between the height systems, called geoid
undulation, can vary significantly in high mountain areas.
Intern
Figure
Figure 6
Figure 8.
