synthesis is postponed to a post-elaboration phase. 
3.1 Photogrammetric data processing 
The following outputs were produced from the orientated 
photogrammetric models: 
- wireframe 3d model: all the geometries, and breaklines due to 
the materials and pathologies, were plotted. This is the most 
widespread output for the architectonic use because it allows to 
take out the conventional 2d drawings as prospects, vertical and 
plane sections. A solid model can also be obtained as a subproduct 
of the wireframe plotting. It requires a large interpolation of data 
and the result is, in any case, a semplified model of the main 
geometries in which the details related to the irregularity of the 
materials can be later added as photorealistic texture; 
- vertical sections, with a step of 10 cm, aimed at making elements 
of comparison for laser scanning data: the sections, indeed, have 
plotted points dense enough and, in addition, derive from an 
accurate selection of more significative discontinuities; 
- digital model: starting from a stereoscopic couple of images with 
the fitting in of the plotted sections as constraint for the altimetry, 
an automatic DEM by image correlation was produced. In order 
to compare the results with laser scanning data, a step of 1 cm has 
been set as a resolution. Output data was interpolated and 
overlapped to the photogrammetric plotting so as to calculate 
surfaces with triangular and square meshes. 
3.2 Laser scanning data processing 
g g 
Laser scanner acquired points are generally more than necessary 
to obtain a meaningful shape representation. At any rate, a kind 
of interpolation from measured data was performed when we 
pass from a discrete (cloud of points) description to a continuous 
one (surfaces). Data processing is mainly based on reducing cloud 
point data, by filtering, and on mesh optimisation. This is like 
saying that, in front of a great number of acquired points, only a 
percentage of them will belong to the final model. 
Presently, 3D modeling of real free-form shapes consists of the 
following steps: registration, pre-processing, mesh generation, 
post-processing, texturing. The range map registration was carried 
out with Cyclone software (by Cyra). 
3.2.1 Registration: Range maps registration were carried out 
with Cyclone software (by Cyra): 20 in all, 13 for the right zone 
and 7 for the left zone. In order to align the range maps, different 
kinds of constraints were used: 
- automatic pre-alignment by specific targets in the lower part of 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004 
Figure 2. 3D model elaboration using topographic 
and photogrammetric data 
the transept; 
- manual pre-alignment by selected homologous couples of points; 
- optimization by the analysis of the overlapping portions of the 
clouds. 
The results are affected by the percentage of overlapping between 
the range maps, by the morphology of the acquired portion of 
the object (better alignment with “strongly 3d” form), by the 
percentage of points used in the computation (3 or 5 % in almost 
all range maps except for the high part of the transept where the 
planarity of the surfaces and the difficulty to collect data 
overlapped enough required from 30 to 50 96 of the points). In 
order to set a unique reference system for all types of data, also 
to compare the results with topographic and photogrammetric 
survey, specific targets, topographically measured, were used as 
constraints in the global range data alignment. The registrations 
were performed in scan block; all of the following data processing 
was done on portions of the entire model, to be able to manage 
acceptable file dimensions and to reduce the elaboration time. 
Also the visualisation became modifiable in real time. At the 
end, all of the blocks were reassembled according to the 
established reference system in order to create a unique model. 
  
  
  
  
  
    
Figure 3. The same range map detail 
before and after noise reduction 
    
  
  
  
    
   
   
   
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
   
   
    
  
    
   
    
   
   
     
    
    
    
    
    
   
    
   
    
   
     
     
  
  
    
  
  
  
  
  
  
  
   
   
   
     
Inter) 
  
Afte 
lowe 
relat 
3.2.2 
to re 
avoi 
the y 
obst: 
Surv 
is on 
lot o 
resu! 
The 
mar} 
one 
and 
data 
mod 
the € 
of pi 
Nois 
by a 
met] 
then 
of tl 
max 
Two 
with 
orga 
surf. 
oper 
redu 
Man 
and | 
Test 
the 
reso 
lot ¢ 
exte