-56- 
all the measured points are transformed in the external reference 
system using the equation (1). 
4.6 Automatic registration 
Using two overlapping scans, the LSR software performs the 
automatic registration following the procedure described in 
paragraphs 2.2 and 2.3. 
The operator can evaluate the success of the registration using 
an analysis of the statistics of the estimated transformation 
parameters and the residuals on the used reflecting targets. A 
first practical evaluation of the procedure can be performed by 
looking at the graphical result of the registration. 
Figure 14. Registration results 
A complete test of the LRS software is described in the next 
paragraphs. 
5. THE 3D MODEL OF A ROMAN BRIDGE 
The object of the survey is to obtain a 3D model of the Pont 
Saint Martin Roman bridge located in the Aosta Valley -Italy 
(see fig. 15). 
Figure 15. The Pont Saint Martin Roman bridge 
Before the laser scanner acquisition, 5 reflecting targets were 
placed on the two sides of the bridge (see fig. 3). The Riegl 
LMS-Z210 laser scanner has been used. This instrument, 
recently acquired by The Politecnico di Torino, has a measuring 
range of 350 m and a measurement accuracy of ± 25 mm. An 
angle step-width range was selected from 0.080 gon to 0.4 gon 
while an angle readout accuracy of 0.04 gon for the line scan 
mode and 0.02 gon for the frame scan were selected. 
Figure 16 shows the location of the instrument during the 
acquisition phase. The acquisition distances range from 3 m and 
330 m; the measured bridge points therefore have an average 
step of about 5 cm. The overlap between two adjacent scans is 
almost of 50 %. 
The selected angle step-width is of 0.080 gon for each scan. 
All the acquired 3D models have been processed using the LRS 
software. All the points of the bridge show a reflectivity that is 
lower than 100 Digital Number (reflectivity range 0-255 DN) 
and the reflecting targets have a mean reflectivity of about 200., 
The program found all the reflecting targets recorded by each 
scan because of the high differences in reflectivity values of the 
Figure 16. Laser scanner acquisition 
x 
Y 
Z 
- 17.451 
8.852 
1.662 
- 39.094 
5.009 
3.381 
- 41.224 
3.648 
4.461 
- 41.146 
3.514 
4.613 
- 30.765 
- 0.125 
1.665 
- 44.041 
8.469 
3.701 
- 43.676 
- 9.286 
3.699 
- 44.811 
- 12.230 
3.951 
- 42.152 
- 28.748 
11.226 
- 34.057 
- 26.667 
5.534 
- 17.459 
- 29.349 
- 1.812 
- 13.792 
- 42.503 
- 0.065 
14.231 
- 36.998 
12.468 
14.111 
- 36.715 
12.288 
Figure 17. Reflecting target extraction 
After the computation of the registration parameters, each scan 
was filtered in order to remove the acquisition noise. As 
mentioned in paragraph 3 this procedure also allows the 
elimination of any acquired points that are not on the object (see 
fig. 18).