aspects
ral the
SUTLS
sue for
as 3D-
in this
logical
ie well-
low the
es with
canning
s — the
surface,
models
InSAR
a small
ground
'cniques
., 2001)
urrently
kind of
omalous
rrestrial
ographic
from a
ugh the
SM of it
ological
ible path
by other
stigation
ric and
egistered
eometric
used for
could be
ncerning
practical
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
applications would like to better focus on problems and
advantages of laser scanning technique applied to geological
sites survey and monitoring.
2. LASER SCANNING SURVEY OF A GEOLOGICAL
SITE
The acquisition of the point-cloud describing the surface of a
slope which has to be investigated is carried out by a long-range
TLS from a single or from multiple stand-points. All scans
acquired from different positions must be fused together,
transforming their coordinates into a unique reference system
(registration). Very often surveys of geological sites require to
be inserted into a pre-defined ground reference system (GRS),
in order to compare data acquired at different times
(monitoring) or for the georeference into a cartographic system.
The setup of the survey's workflow stricly depends on
characteristics of TLS, registration technique, dimensions of the
site, required data accuracy and resolution. In the following
some of these aspect will be analyzed in detail.
2.1 State-of-the-art of long-range TLS
A long-range TLS (LRTLS) should allow the acquisition of
points belonging to a surface at a distance of several hundreds
metres from the stationing position. Upgraded technical features
of current available instruments can be found at websites of
constructors (see References from Websites).
The common measurement principle of all TLS is the time-of-
flight method, the unique which permits long range
measurement in a fast time. More recent instruments are
equipped by a digital camera as well, which may collect images
to be used for generating orthophotos or realistic Virtual Reality
models.
Almost all current LRTLSs present some facilities which can be
very helpful in practical surveys, such as the possibility of
controlling the acquisition process by means of a laptop or a
palm PC, also via remote wireless connection.
From a logistic point of view, in geological surveys the
transportability of the instrument and its accessories (tripod,
energy unit, PC) is fundamental, because might easily happen
that arduous stand-points have to be reach. However, many
efforts are required in the future to reduce weight and
dimensions of LRTLSs.
2.2 Registration of views: different approaches
Concerning strategies for registering multiple scans, two
different approaches can be followed. In both cases, a set of
ground control points (GCPs) is needed to register the point-
cloud to a given GRS. As GCPs may be used refro-reflective
targets or common features which are well identifiable in the
scans. Which method is available strictly depends on the TLS
and particularly on the data processing software being used.
Normally only one option is possible.
2.2.10 Fully GCP based registration: In case enough GCPs
are available in each scan, these can be straight-forward
registered to the given GRS. From a mathematical concern, this
task consist in computing 6 parameter of a rigid 3D roto-
translation from the intrinsic reference system (IRS) of each
3D-view to the GRS.
709
Although this method is largely reliable and leads to a high
accuracy in the registration process, sites which are surveyed
for geological purposes may present a very complex
morphology, which makes difficult, if not impossible, to carry
out topographic measurements of GCPs. This drawback may
results in a longer survey times and decreases the avantage of
using a TLS in many applications.
On the other hand, if different 3D-views share a sufficient
portion, GPCs can be positioned to serve more than one model.
Moreover, if the TLS features a large horizontal FoV, GCPs can
be also external to the area to be acquired.
Usually, the minimum number of GCPs needed to register a
3D-model is 3, but practically a higher one may be required,
according to the algorithm used to compute the initial values for
the roto-translation parameters (equations are not linear); in
literature a large variety of methods to solve for these
approximations can be found (see Gruen & Akca, 2004).
2.2.2 Pairwise registration techniques: A second
possibility of georeferencing different 3D-views arises in case
GPSs are not visible in all of them. In this case the available
methods are based on the pairwise registration of scans, starting
from a 3D-view which is chosen as reference and registering to
it all the other 3D-views sharing a sufficient number of tie
points (TPs). Then, neighbouring scans are registered as far as
the whole block is oriented. As tie points, in geological
applications retro-reflective targets have to be preferably used,
because their measurement can be performed with the higher
accuracy either by automatic and manual procedures. Methods
based on automatic extraction of not-signalized tie points (see
Gruen & Akca, 2004 for a review) or directly on matching
corrisponding surfaces are very usefull in architectural and
mechanical applications, but they do not hold yet in the
geological field, where the presence of not coherent terrain and
of moveable objects (rocks, vegetations, ...) can lead to errors in
registration. After pairwise registration of all 3D-views to the
same “project” reference system (PRS), thank to a set of small
GCPs the point-cloud can be transformed into the GRS.
In Scaioni & Forlani (2003) a procedure has been proposed to
solve for a simultaneous block triangulation of all models, using
as constraint only a small set of GCPs. This approach (or
another similar) can be usefull if in the interested area GCPs
cannot be directly measured, being only possible to put TPs.
2.3 Survey planning
The planning of a landslide survey usually suffers from the
typical problems involved in a traditional photogrammetric
survey: several factors must be considered, so that standard
design methods are not so easy to be established, as happens for
flight plans in aerial photogrammetry. Nevertheless, a limited
effort has been produced so far in order to give at least some
basic addresses for a correct survey planning.
The strategy we followed is based on establishing a set of
simplified relations which can be used to setup a TLS survey.
Topics decision that have to be made are the following:
e selection of the more suitable TLS model (if possible);
® position and number of TLS stations;
e strategy for 3D-views registration, resulting in
positiong and measurement of GCPs.
Input data for the survey design are basically the resolution and
the accuracy of the point-cloud describing the surface to
