International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
  
Figure 6. Classification: non terrain points (vegetation and other 
features) in green, over terrain surface in brown. 
5. FINAL POINT CLOUD GEOREFERENCING 
After the data capture, orientation, merging and classification of 
the two point clouds, we proceed to data transformation to a 
reference system global and common to both campaigns, in 
which surface and terrain models can be compared to study the 
landslide process. 
Using as common points in both reference systems (local 
system used in scanning process and global ERTS89 system) 
the positions of GPS antenna phase center incorporated to point 
clouds, a rigid 3D transformation of six parameters is applied to 
the integration of point clouds corresponding to both campaigns 
in a common reference system in which they can be compared. 
However, some cautions have to be taken in account because of 
these points are affected by the errors in the orientation process 
of single scans carried out by means of surface matching, and 
the further propagation of those errors. 
To solve these problems a weighting of these points taking in 
account its quality is carried out to calculation of transformation 
parameters. In this way, the 3D transformation was made in two 
steps: first, we made a transformation in which every point has 
the same weight; after that initial adjustment, we recalculate the 
transformation taking in account the first results (the better 
adjustment of one point in the first transformation, the higher 
assigned weight in the second transformation). These deviations 
obtained for the initial transformation used in the second are 
shown in table 1; in this way, we obtain more reliable 
parameters than in a non-weighted transformation. 
Finally, this type transformation is applied to point clouds 
corresponding to both campaigns, so all data (point clouds and 
surfaces obtained from them) are now in the same reference 
system (ERTS) that will be also used to data integration of 
further works. 
The errors found in deviations (table 1) are very similar to the 
trend of those obtained for every scan in the phase of relative 
orientation by means surface matching. This similar trend 
indicates that the correlation errors could be used to weight the 
corresponding points in the 3D transformation. However, these 
correlation errors between different scans can be affected by 
several factors (presence of vegetation, size and points density 
of overlapped areas between the scans to be correlated, terrain 
surface morphology, scan station position, etc.), so errors could 
not be homogeneous. In this way, a weighted transformation 
24 
based on the deviations obtained as we explain before can give 
more reliable results. 
  
  
Scan-stations Vx Vy Vz 
11 -0.0084 0.0183 -0.0365 
12 -0.0075 -0.0093 -0.0569 
13 -0.0042 -0.0309 -0.0022 
14 -0.0248 -0.0600  -0.0013 
21 0.0383 0.0284 0.0857 
31 0.0298 0.0449 -0.0732 
32 0.0330 0.0426 -0.0428 
41 -0.0236 0.0368 0.0295 
42 -0.0222 0.0095 0.0184 
43 -0.0016 0.0109 0.0308 
44 -0.0001 -0.0205 0.0288 
45 -0.0085 -0.0375 0.0034 
46 -0.0004  -0.0333 0.0163 
11 -0.0084 0.0183 -0.0365 
12 -0.0075 -0.0093 -0.0569 
  
  
  
  
Table 1. Obtained deviations in the initial transformation of the 
scanning points corresponding to the second campaign. 
6. FIRST RESULTS AND DISCUSSION 
Once data are in a stable and common reference system we can 
made the comparative analysis working with point clouds or 
with derivate surfaces. 
In this way, we first observe two rupture zones in the road cut 
affecting also to the upslope (figure 7); they were generated as a 
consequence of the instability processes affecting to the region 
after the heavy rainfall in 2009/10 winter. This unstable zone 
mobilized an important volume of flowing material that invaded 
one of the carriages of the A-4 national highway and produced 
some traffic interruptions and other problems. Besides, the 
landslide evolved between the two considered campaigns; it can 
be clearly observed in a quickly view of both surfaces, with 
important superficial displacements and formation of steps, 
scarps and cracks. 
From digital surface model (DSM), some superficial 
movements of terrain and vegetation are roughly measured by 
means of the tools of I-Site software, obtaining maximum 
displacements of about 8-9 m between both campaigns that 
produces a daily rate of about 0,55-0,65 m day !. 
  
Figure 7. Observed displacements of the terrain and vegetation 
and terrain in unstable zones of the studied road cut. 
Inter 
On the otl 
means of t 
figure 8, 
Depletion 
the seconc 
first camp 
second su 
results to t 
Mobi 
mate 
Volum 
Rates (n 
  
Table 2. 
These cal 
material) : 
between 
because c 
occupying 
road cut 
actuation 
objective 
works sta 
amount « 
consisted 
contentio! 
  
Figure 8. 
in orang 
campaig 
surface i 
center: | 
7 
Terrestri 
useful to 
sense th 
deforma 
spatial a