ul 2004
Its.
a rough
. One of
nique is
hich get
of each
ield, the
jyllection
eduction
ding to
istration
ich may
:ological
t least 4-
he direct
t reliable
tise this
me CPs
ographic
ssible. In
must be
bility of
60°, it is
d-points,
êre GPS
nned at
landslide
hould be
ermanent
not very
jning of
:d, GCPs
‘be not
positions
a strong
juld be to
ble sites
ed in the
'om these
d GCPs
y time by
of targets
ect which
Yet TLS
heir own
, such as
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
cilinders or small spheres), in surveys where long ranges are
involved these may often appear as to small; furthermore, their
topographic measurement may be difficult. In experiences
carried out so far, we adopted planar retro-reflective targets as
TPs, and special targets made up of a retro-reflective disk
mounted on a tripod as GCPs, these being equipped of a system
to fix a prism or a GPS antenna (Fig. 2).
2.4 Data acquisition
This stage is the core of the whole process, because the
complexity of the context to be surveyed requires to verify all
the hypoteses established during the design stage. By
positioning the TLS in all planned stand-points, the complete
coverage of the object at the wanted ground resolution can be
checked; also the acquisition of CPs must be verified. To do
this, a really usefull instrumental tool is the preview at low
resolution of the whole scan. After this check, possible
modification or integration of the survey layout can be carried
out during the same measurement campaign.
Another important task is the verification of the registration
procedure, resulting from a correct targets’ displacement and
acquisition in all scans.
Acquisition of digital images can be carried out during the
range data collection, in order to add up information about the
color texture of the objects (Sgrenzaroli & Wolfart, 2002).
Digital camera may be fixed to the TLS or may be used
independently: in the former case, the image geometry of the
camera can be related to that of the scanner, so that a
corrispondence between each point of the 3D-view and the
image can be established by “on the job calibration”, to be
carried out only once time before an acquisition session (Ullrich
et al., 2003). In the latter case, each image can be oriented with
respect to the point-cloud in post-processing by space resection.
In both cases, image registration to the point cloud is performed
by manually measuring well identifiable points on the 3D-view
and the images. The a priori knowledge of intrinsic camera
calibration parameters would reduce the number of control
points to adopt.
2.5 Data processing
Data processing is based on three main stages:
I. pre-editing of each scan, i.e. resampling of scans in
case they are too dense and measurement of CPs;
2. registration of all 3D-views to a given GRS; cleaning
of points located in not interesting parts, in order to
reduce the total amount of data before next stages.
Furthermore, after registration, large portions of the
point-cloud may be made up by the overlap of more
scans. Thank to a filtering method, duplicate data
should be eliminated;
3. extraction of required information.
In the last stage, information needed by the final users has to be
derived. From an operational point of view, this task (as the
registration and pre-editing as well) should be performed by
devoted SWs, either commercial or scientific. In the following
we briefly address to some products that can be derived from
the registered point-cloud which may result very usefull in
geological investigations and analyses.
711
2.5.1 Digital Surface Model: From the registered point-
cloud a DSM describing the external surface of the surveyed
slope can be derived. Thank to meshing techniques, the set of
raw 3D points is converted into a continuous surface and thus
results in a visually more intuitive representation and in
reducing the amount of data. Moreover, meshing represents a
preliminary task to the extraction of sections and contour lines,
to generation of orthophotos and photo-textured models. The
widespread used meshing technique is /riangulation of scan
data (Edelsbrunner, 2001).
The availability of an accurate DSM of a landslide area may be
very important for geological investigations, for which usually
only rough information read from mid scale maps is used. In
particular we would like to focus on some products than can be
obtained from the DSM:
* as input data for specific SW simulating possible
landslide behaviour;
e simulating the path of possible falling rocks;
* computation of volumes (and their variations if multi-
temporal data were available);
e computation of vector field describing movements of
the landslide surface.
25.2 Cross-ections and countour lines: Very simple
products that can be derived from TLS survey are cross-sections
and horizontal countour lines. Thank to the high density of
points, they give an accurate description of the site, very usefull
in planning of works of consolidation and protection. Because
of the possible presence of moving objects, a smoothing method
should be preferably applied to remove noisy data from
extracted lines.
2.5.3 Photo-realistic 3D Models: Among different
approaches to display 3D models (wireframe models, shaded
moded — see Remondino, 2003), photo-texture 3D models allow
the most realistic visualization of the object. Texture mapping
involves an image being mapped onto the surface composed by
a triangulated model, so that the colour of the object at each
pixel is modified by the corrisponding colour derived from the
texture (Sgrenzaroli & Wolfart, 2002).
2.5.. Orthophotos: The knowledge of a DSM allows to
generate orthophotos, that may be very usefull in geological
analysis. Indeed, orthoimages offer a detailed view of the slope,
which could be intergated by other information (cross-sections,
contour lines, positions of different sensors an so on).
2.5.5 Topographic maps: Nevertheless, information
described so far can be integrated by a vector map directly
derived from the 3D point-cloud. For example, while the DSM
and its by-products are the most appropriate methods to
represent a slope, concerning building, street, infrastructures,
these could be better drawn by extracting their contours.
Moreover, vectorization results in a simplification of the whole
point-cloud, which becomes lighter to be managed.
3. TWO APPLICATIONS
3.1 A landslide in Caslino d’Erba (Como)
The first application concerns the acquisition of 3D model of a
vertical slope threathening a portion of the village of Caslino
d'Erba, located between towns of Como and Lecco in Northern
