Istanbul 2004 
merge images 
gles, or even 
tions between 
are applied to 
'ormance, the 
developed at 
| tools. Scene 
organize and 
e data. They 
ics rendering 
LING 
he interior of 
S-Z360 laser 
the room and 
G scans were 
alls and 2 for 
nd by placing 
ial resolution 
was obtained. 
idle with the 
ion, both this 
for the 3D 
Modeler of 
/ery powerful 
bjects, whose 
dense point 
ch allow the 
. control over 
sults through 
ving modules 
ocedure was 
ceiling scans 
ly helped the 
itching. points 
approximate 
ined and then 
- the refined 
thm. In both 
the reference 
(d alignment 
5 of the first 
ielded a very 
e of 6.3 x 10 
id an average 
result should 
| procedure 
s due to the 
lly, through 
to model the 
| is the most 
re 6). 
  
  
Peemeters Sinisbcs |Comperisee] - 
WSerstons 
Coréréénes 
Inch Cory Mées SidDaw 
1 &3e-008 9003323 2004384 
  
  
  
  
5.3e-008 0004072 2006918 
36-008 900409 004120 
6.3=-108 OUZ416 DRISERN 
3e-008 00605104 079 
Tz-005 SOISHAN D008 
  
   
Les 294 La Lu À 
i5 gn 
  
  
  
  
  
  
  
Figure 6. The meshed 3D model from range data, 
rendered as a unique point cloud. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004 
6. COMPARING THE MODELS 
In order to compare the two models (image-based and laser 
scanning), the aligned point cloud was simplified (i.e. 
decimated), deleting most of the overlapping points. This 
approach has two advantages: 1) it allows to reduce the model 
size in terms of number of points, resulting in a small file size; 
2) overlapping points lead to a redundant information that is 
only useful for the alignment and the subsequent triangulation 
step. not for the model georeferencing. It should be noted that 
both the georeferencing and the comparison stages has been 
performed only on the decimated point cloud making up the 
laser scanner 3D model. The triangulated model was not used 
here: due to the inherent “nature” of the meshing process, 
points are substituted by triangles and further simplifications of 
the cloud can occur, i.e. points can be discarded leading to a 
change respect with the original data. 
To compare the two models, they need to be registered in the 
same reference system, here defined by the total station survey. 
Therefore the control points measured with the total station 
were identified, by visual inspection of the operator, both on 
the 3D laser and on the image-based model. As both natural 
and artificial targets were used, the comparison has been split 
in two different parts accordingly: in a first run we employed 
the natural features only (points selected on the frescoes) while 
in the second run the retroreflective targets were taken into 
account. This approach was adopted to check the reliability of 
a laser scanning model georeferencing using natural targets 
only, which are the unique kind of targets often available. 
Twenty-eight of the 50 surveyed natural targets were used as 
control points for the model georeferencing while the others 
were considered as check points to evaluate residual 
differences between the two models. This step was very time 
consuming, because those points had to be carefully selected in 
order to be not coincident with the ones identified as check 
points on the laser scanning 3D model. Moreover, two further 
issues ensued during this processing phase: 1) the low quality 
of registered intensity data made very difficult or even 
impossible a reliable recognition of the control points measured 
with the total station; 2) the retroreflective targets saturated the 
laser sensor, resulting in a set of undefined white spots of 
points, of no use for the comparison with the image-based 
model. For those targets, the points most probably measured by 
the total station were identified as the intersection of the two 
diagonals of the rectangular spots, and then used for the model 
georeferencing only. This stage was accomplished through the 
powerful tools available in the /m/nspect module of Polyworks 
software. Basically, only 22 features on the frescoes could be 
well recognized and used as check points and 15 artificial 
targets out of 20 could be employed for the laser scanner model 
georeferencing. 
An analysis of the comparison results using the 22 check points 
for both models is reported in Table 3, where RMS values 
denote the differences between coordinates of check points 
selected on a 3D model and corresponding points, measured by 
total station. 
Table 3. Comparison between models 
  
  
  
  
3D Model # of Check RMS X RMS Y | RMS 
pts [m] [m] Z [m] 
Image-based 22 0.017 0.025 0.020 
Laser scanner 22 0.056 0.063 0.044