1263
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part Bl. Beijing 2008
Figure 14. 3D View in GIS of DSM and Orthphoto
3.4 Randa, Switzerland
The goal of this project is the characterization of the May 1991
Randa rockslide failure (Figure 15) surface using UAV
photogrammetry. The project is embedded into the project
"Progressive development of shear/slide surfaces in rock slopes
(Phase III)" of the Engineering Geology group at ETH Zurich
(Randa, 2008 and Eisenbeiss, 2008).
Figure 15. Randa rockslide
Data collection of this project did primarily involve a detailed
investigation of the visible failure surfaces and back scarps of
the 1991 events at Randa. Although an important part of the
sliding surfaces is covered by debris, the visible parts of the
detachment surface are large and include both secondary sliding
planes as well as lateral and back scarps showing in detail the
interactions between all fracture sets, faults and rock bridges.
These interactions are critical for the evolution of rock slope
instability. Given the size, the steepness and the dangerous
access to these outcrops, fracture mapping will be performed
from inspection of high-resolution digital surface models
(DSMs) and stereo photographs of the slope.
The resolution of the generated rock surface models allow for
the recognition of planes and traces describing fractures and
faults contributing to the 1991 slope failures. It also permits
accurate measurements of fracture and fault orientation and
trace length. To meet these requirements, the DSMs were
constructed from high-resolution images taken from locations
close to the rock outcrops whose position is known with a
sufficient degree of accuracy. In the framework of this project,
digital photographs of the 1991 failure surface were taken from
a UAV which is a suitable technology for imaging inaccessible
outcrops with a high resolution.
Photographs of the rocky outcrops were used to derive a dense
DSM with 10-20 cm spatial resolution, and a height accuracy of
1-2 cm, allowing for detailed morphometric analysis and
characterization of fracture surfaces. To achieve a height
accuracy of 1-2 cm, a large image scale was required involving
the acquisition of many images and control points. Furthermore,
the control points used to orientate the images should also be
very accurate, better than 1 cm. For the object coordinates of
these points this accuracy can be achieved, even if they are
natural points, by measuring them geodetically with a total
station. However, access to the site is limited. Therefore the
control points could only be installed at the border of the site.
DSMs were produced automatically using advanced multi
image matching techniques, and orthoimages were generated.
The orthoimages are used for 3D visualisation and visual
interpretations. The first results from a test flight in May 2007
are given in Figure 16. The left image represents a zoom-in of
one of the acquired photos. The right image is the derived
DSM of Randa, showing 50 times higher resolution than the
existing lidar dataset of the site.
Figure 16. Snapshot of Randa Photo (left) and DSM (right)
4. DSM COMPARISON
The authors have been using SAT-PP and Match-T for accurate
DSM production. While SAT-PP was used for the maize field
and Randa project, the highway and the Geofort were processed
with Match-T. Both software packages are based on a matching
approach using a coarse-to-fme hierarchical solution with a
combination of multi image matching algorithms and automatic
quality control. The software matches three kinds of features -
feature points, grid points and edges on the original resolution -
progressively starting from low-density features on the low
resolution level of the image pyramid. A TIN of the DSM is
reconstructed from the matched features on each level of the
image pyramid by using the constrained Delauney triangulation
method. This TIN in turn is used in the subsequent pyramid
level for approximations and adaptive computation of the
