. Istanbul 2004 
te 19th century 
mages due to 
ces to crumble 
diately and, on 
IS Set up, a sort 
d environment 
id monitors its 
these unique 
id compare 3D 
lata sets, four 
for this work: 
3D and Cyrax 
érformance in 
mum distance, 
nary of their 
ng hours to be 
riate positions 
lifferent times, 
nner from the 
eeds provided 
re collected in 
could not be 
r, therefore we 
patches of the 
data arc from 
and 5% from 
ution and the 
' next section. 
oifications 
  
Riegl USA 
  
LAS Z2Y0i 
MÀ 
  
904 
  
im 
1 
  
  
50 mmm at Sí m 
  
liDAR 
  
&8ü0O0U 
  
12,000 
  
15 mm at 400 m 
  
6 mm at 100 m 
2 seconds 
  
y 
* 
4m 
  
500 m 
  
  
  
  
Table 2. Cyrax and Optech laser scanner specifications 
MANUFACTURER 
  
Cyra Technologies || Optech Inc. 
  
  
  
  
  
  
  
  
  
PRODUCT HDS2500 ILRIS-3D 
PERFORMANCE 
Laser Wavelength (in nm) 522 1540 
Laser Power (in W, mW) « | mW avg 10 nàV 
FDA Laser Classification (Class) 2 Cas var gerne mes 
Beam Diameter at Specified Distance x6 mm from € | 0675 in at lon 
(0.Y ft at X ft/Ymm at X m) 50m 1/17 mm at 30 m 
Measurement Téchnique Time of flight | LDA (Time of Flight 
Average Data Acquisition Rate (pps) 1,000 2,009 
.Maximum Data Acquisition Rate (pps) 1,000 3,000 
  
Distance Accuracy at Specified Distance 
(O.Y ft at X ft/Ymm at X m) 
Position Accuracy at Specified Distance 
(0.Y ft at X ft/Ymm at X m) 
Angular Accuracy 
0.275 in of 330 1/7 
mm at 100m, [2] 
14 in at 330 ft/10 
mm at 100 m, [2] 
€0 micro-radians | (9 eges Caco 3 
4 mm at 50m 
  
4 mm at 50m 
  
  
  
  
  
  
  
Minimum Range (feet/m) 15 m 10/5 m 
Maximum Range (feet/m) 100 m NEO R> 1,506 pic] 
Field of View (vertical angle) 40 degrees [8j 
Field of View (horizontal angle) 40 degrees | Pen Asp 
GENERAL 
Scanner Dimensions (LxWxH) hia xan {1200 IIS BIE] 
  
  
  
  
  
Scanner Weight (pounds/kg) 20.5 kg 25 Ibs/12 ke 
  
4. THE 3D MODELING 
As previously mentioned, the Scrovegni Chapel was chosen for 
this project because of its simple architecture: the presence of a 
single hall (nave) allowed us to survey the interior with case 
without occlusions of the walls and the vaulted ceiling. 
Conversely, the apse has revealed to be more tricky to be 
completely surveyed given the presence of the main altar, 
which made impossible to scan some parts of the walls, given 
the small room available between the back of the altar and the 
apse. This area resulted in a wide hole in the 3D model that had 
to be manually closed using both artificial surface patches and 
a few digital images as reference guide for the hole filling 
stage. 
Since all employed laser sensors were able to acquire the 
intensity of the reflected beam, figures of the paints were 
imaged on the point clouds, as well. Such features could be 
often well recognized on the intensity data and therefore 
extensively used as reference points during the alignment 
procedure, as described in following subsection. 
Prior to registering the scans, the range data were linearly 
interpolated in order to get a uniform point spacing, since they 
were acquired with different resolutions: 8mm for Mensi and 
Optech, 35mm for the Cyrax and ! em for Riegl. Accordingly a 
8mm grid size was chosen as a trade-off between point cloud 
density and size of related data files being processed. However 
the interpolation was applied to Cyrax data only, while Riegl 
scans were left unchanged to avoid the addition of artificial 
(i.e. not real) points in the corresponding clouds. All the 
processing steps were carried using Polyworks [Innovmetric 
inc.] a powerful 3D modeling software which allows to work 
with multiresolution range data. 
Then a 2 steps interactive manual N-points alignment 
procedure was adopted to register the scans with each other: 
matching points were easily recognized using only the intensity 
data, as shown in figure 3. As most part of the church was 
surveyed by the Mensi laser scanner, (95% of the hall and 10% 
of the apse), related scans were used as main data block for the 
registration, at this stage. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004 
  
Figure 3. Interactive manual alignment. 
After all the scans were manually aligned, a global ICP-based 
registration algorithm was applied in order to refine the results 
of previous step. Such approach [Soucy et al, 1996] yielded a 
very good registration for the Mensi scans, with an average 
RMS alignment error of 0.006 m, confirming the goodness of 
the registration procedure implemented in Polyworks: the 
residual error is due to the inherent accuracy of the laser 
scanner. However, if remaining scans of the Cyrax, Optech and 
Riegl laser scanners are taken into account, then the RMS 
grows to about lcm, as showed in figure 4. This increase could 
be explained considering that, among employed laser sensors, 
the Riegl was the noisiest and related scans were used to join 
the range data from the nave (surveyed mainly with the Mensi) 
with the ones from the apse (completely surveyed by the Cyrax 
only). 
An example of the comparison of noise content between range 
data of Riegl, Optech and Mensi laser scanners is presented in 
figures 6 b-d, which relate to the survey of the surface of the 
wall displayed in figure 6a. 
  
*; IMAlign - Alignment &t 
Parameters Statistics | Comparison 
# lterations 
  
33 
Convergence 
indx Conv Mean .StdDev 
  
  
46 3.0e-008 0.529448 10.14501 a 
47 30e-006 2404396 6.263409 
48 3.0e-008-0.356729 7.671836 
49 3.0e-008-0613273 12.56912 
50 3.0e-008 0.353304 12.53452 
51 3.0e-008 1.313603 10.47731 
52 3.0e-008 1.939209 9.818244 
53 3.0e-008 0.840151 11 27706 
54 8.2e-010 0.316092 8.916086 
55 82e-010-0.218566 11.26402 
56 &Ce-010 0.180667 10.791569 
5? 82e-010-0 252367 4242511 
58 7.0e-011 0.777804 5.659035 
59 7.0e-011 0.545312 7.084582 x] 
c^ 010 ^43 PARAM AR TAC An Ar 
Start | Close 
Figure 4. Results of global alignment