hul 2004 
tion 
regular 
manual 
ation of 
tion and 
e, is an 
ssible to 
trips" of 
sections 
lines of 
ure that 
needs to 
he given 
orations 
g to the 
order to 
eractive 
Opposite 
ount the 
stric and 
ation of 
riangle) 
es; 
y 
>dge 
) with 
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
  
fm] 
A a Table 3. 
8019 4-—————————————————Wa- Comparison between 
0,008 L ee otii M)... targets, topograptu- 
| cally measured, and 
0,006 | 
the same ones on the 
cloud of points after 
global registration 
0,004 + 
0,002 + 
    
0,000 
  
  
  
  
    
2o wy 
«o sa 
e 
) = 
  
euo e 
n G 9% SF ua 
om m 
Table 4. Comparison between natural points, topographically 
measured, and the same ones on the surface model 
  
  
  
  
  
  
  
  
  
  
  
Figure 10. Graphic comparison between photogrammetric 
sections and laser scanning model. Residuals < + 2 cm 
the polygonal model can be carried out with criteria similar to 
those adopted to decimate the cloud of points. In order to fulfil 
the aforesaid need to maintain, as well as possible, the geometric 
shape, it is possible to choose to leave the details related to the 
material texture. Before applying the polygonal decimation, 
breaklines may be detected. These breaklines are useful for the 
edge reconstruction and can be computed as fixed boundaries in 
à stronger polygonal decimation. Surface smoothing can also be 
applied with the aim to increase the quality of the mesh, 
optimizing its geometry, without modifying fixed boundaries. 
4. METRIC EVALUATION 
l'he accuracy of the different kinds of models obtained from la- 
  
  
  
  
     
  
   
  
   
  
   
   
     
   
   
  
  
   
   
  
  
„101 x] 
Dati Fotogrammetrici I Leggi dati Laser ] Dati a confronto 
: i | 
] ] | Zim ]*Y calcolato | Tipo = Caicola | 
E35] j C 109 3431 —Interpolszione eof 
iE (* Valoipesal (^ Minimi quadrati | 
1 Ampiezza intervallo dei punti laser da ] 
Te considerare: 2D m H 
ET E =] 002 m 
6 2 ; 
2] 2 | NOTÁ: se non vi sono valori laser 
8 : B nell'intervallo, viene posto il valore 
13. 5B 100000 
10 zB v 
11 ?; B Visualizza T 
12 ? B Jv Dati fotogrammentici 
13 ? B ivi i 
14 ? B Iv Dati calcolati 
15 78 2 pum MU 
16 108,2801 2 à Salva Risultati | 
17 108.2842 . 108,9209 2 B 
18 108,7 1082811 108.918 1 Ev | 
  
  
  
  
  
  
Figure 11. Screenshot of the program for comparison 
between photogrammetric section and the “strips” of points 
extracted from the cloud: Residuals < £1 cm 
ser scanning does not depend only on the acquisition accuracy 
(type of instrument or scan resolution) but also on the different 
data processing, applied both on the cloud of points and on the 
triangulated surface. At the end of the main elaborations of the 
acquired data, a comparison between the final resulting mesh 
and the original acquired cloud of points was carried out. For 
that purpose sample tests have been done both on punctual ele- 
ments and on tridimensional surface with the statistic evaluation 
of the standard deviation. 
4.1 Measured targets vs. automatic recognized targets on the 
cloud of points 
As mentioned above, the registrations were performed in scan 
block. Every block is referred to the unique reference system 
thanks to the special targets, topographically determined, and au- 
tomatically recognized in the cloud of points. There is no inter- 
pretation or collimation error to determine them. The residuals 
are of the same order of magnitute as measurement accuracy of 
the employed instruments (= 6 mm) 
4,2 Measured natural points vs. manual extracted points 
on model of surfaces by laser scanning 
Another comparison, point vs. point, was performed between the 
complete model of surfaces and the topographic measure of the 
detail points. In this case the points on the model were manually 
chosen so they are affected by an unavoidable interpretative com- 
ponent. Indeed it is not possible to separate the metric evaluation 
from the descriptive one. Thirty natural points, well distributed 
on the selected area, were measured with reiterated collimations: 
the mean residuals are not much higher than one centimeter. 
4.3 Model laser scanning vs. photogrammetric plotting 
sections 
Vertical sections, every 10 centimeters, were plotted with a Digicart 
analytic plotter and overlapped on a triangulated model: the 
average distance between the plotted vertices and the surface is of 
1.3 em, with standard deviation of 0.9 em. 
The» visualization of these distances by means of a colour map 
allow to highlight the distribution, the systematic error or the 
deformation of the model. 
4.4 Laser scanning sections vs. photogrammetric plotting 
sections