International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
  
  
Figure 7. Partial reconstruction of the holes in the low part of the bass-relief: lion meshes (left: segmentation; right: reconstruction). 
4. DATA ELABORATION 
The first operation is the data decimation (Figure 9) and it is 
the most time consuming, also using special frames to delete 
points which are in or out the frame. Some patches can be 
created, using a defined layer where to interpolate some 
points; after the layer is removed with a simple operation 
(on/off). Also, a uniform decimation is possible when the 
data points density seems to be very high in that context. 
In the Table 8, some results are reported about the 
decimation phase (DEC), where the number of points before 
and after decimation is reported. 
  
  
  
  
  
  
  
  
Scanworld | # before DEC | #after DEC % removed 
1 3.590.766 2.780.590 22,694 
2 3.832.514 3.257.432 13% 
3 402.780 273.292 32% 
4 432.376 313.332 27% 
5 259.573 233.603 9,2% 
6 2.531.445 2.101.389 17% 
  
  
  
  
  
Table 8. Decimation phase and % of removed points. 
ARE Br 5 apa i A 
Figure 9. Example of decimated view (Scanworld 4). 
  
The second phase of the data elaboration is the so called 
Registration. This is a mosaic process where more Scanworld 
are "registrated" in the same coordinate system, using 
couples of homologous points. 
Two kinds of registration have been used for this acquisition: 
- the target registration (point to point or sphere to sphere) 
- the cloud point registration (cloud to cloud). 
In the first case, the Cyra Tie-point and natural points (well 
visible architectural features) have been used. In the second 
case, the target aren't needed and two Scanworld must be 
aligned using some points chosen by the user. 
An example of alignment optimization is shown in Figure 10, 
including the related error histogram at the first iteration 
  
STEP 3: optimization of the alignment - i? iteration - 
  
  
   
  
  
  
  
  
Figure 10. Example of alignment optimization. 
5. CONCLUSIONS 
Different registration procedures are used for the laser 
scanner survey of the monumental door called "Portello" in 
Padua. Also the CloudConstrainWizard has been used to 
create some cloud constraint in order to register more 
Scanworld. In the same time the minimization of single 
procedures is performed. Finally, the global registration can 
be operated. In this survey the error, after the global 
registration, appears quite good in the range of | em; in only 
one case this value is exceeded. Probably, this fact is due to a 
lightly different acquisition mode for the lateral prospects. 
The Table 11 reports the results of the final alignment with 
   
  
  
   
  
  
   
  
  
   
   
   
   
   
   
   
  
   
   
  
   
  
   
   
    
  
   
  
  
  
    
     
  
   
  
  
  
  
  
  
  
  
  
        
  
Internati 
the asso 
and Max 
The Fig 
the glob 
outline c 
  
Figure 1 
6.1 Reg 
Pulli K 
in 3D 
proceed 
Besl P. 
shapes, 
Machir 
Potmes 
match- 
Conf. « 
CA:Co 
1089-1 
El-Hak 
D .Iw 
virtual 
in Pr 
Sensin,