and shutter speed settings were adjusted to obtain optimal target 
exposure. Twenty-eight targets affixed to three steel beams 
acted as stable reference points for datum definition. These 
were bolted to a scaffold structure that was erected independent 
of the bridge. 
4.4 Photogrammetric Deformation Analysis 
Image point measurement was made using Australis (Fraser and 
Edmundson, 2000) and free network bundle adjustments were 
performed using the FEMBUN software (Lichti and Chapman, 
1997). Points on the steel beams were used to define the datum. 
Pre-calibrated interior orientation parameters (principal 
distance, principal point, radial lens distortion and decentring 
distortion coefficients) were applied as constants. Object point 
precision in the height or Z dimension—the most pertinent for 
subsequent structural analysis—was approximately ±0.4 mm for 
the non-datum points. The 2-minute data acquisition time 
constraint precluded a realisation of greater precision by 
capturing more images. 
Since all epochs of imagery possessed the same datum 
definition, deformation analysis was a matter of subtracting 
loaded co-ordinates from no-load co-ordinates. Figure 6 shows 
the result of one such comparison. The vectors indicate 
deflection of the four longitudinal stringers (3-1, 3-2, 3-3 and 3- 
4) in response to a load of 60.65 t placed at mid-span. As 
expected, the maximum deflection (-8.06 mm) occurred at mid 
span, and defection decreased near the ends where the stringer 
was supported. 
4.5 Scanner Deformation Analysis 
Stringer 3-4 was also scanned with the I-SiTE to further 
ascertain the sensitivity of the scanner for deformation 
measurement. The scanner remained static throughout the 
testing. Due to the 2-minute data acquisition window, only a 
single scan could be captured at each load increment. In light of 
the results in Section 3 and the small deformations on Stringer 
3-4 (-3.3 to -5.2 mm) under the 60.65 t load, expectations about 
the ability to accurately recover the deflections were low. 
The analysis procedure applied for the cylinder testing was 
utilised for the stringers. Scan clouds were edited to remove 
spurious returns and then -modelled with the “triangulated 
surface with second-order least squares trending” (Maptek, 
2002). Stringer cross-sections were extracted at the locations of 
the photogrammetric targets to facilitate direct comparison of 
displacements. Deflections were estimated from the vertical 
displacement between the no-load and loaded cross-sections at 
both the top and bottom of each section. 
Numerical deflection estimates at eight photogrammetric target 
locations on stringer 3-4 are presented in Table 3. The scanner 
displacements measured from the bottom are clearly biased with 
an RMS error of ±9.1 mm. Displacements measured from the 
top of the cross-sections are much more encouraging, with an 
RMS error of ±4.9 mm. In this case, the errors are nearly 
constant. 
Error sources in the displacements include the scanner 
measurement noise, surface model interpolation error and the 
shape distortion highlighted in Subsection 3.2. The latter is 
believed to be dominant error source, particularly at the bottom 
of the stringers where the shape distortion was clearly evident. 
Rectification of this distortion is the subject of ongoing 
research. 
CONCLUSIONS 
The issues of laser scanner calibration and benchmark testing 
are important for both heritage recording and metrology 
applications in order to assure data quality. A series of rigorous 
tests have been conducted in a lab environment and under real 
conditions in order to quantify scanner performance. In terms 
of precision, lab testing indicated that the scanner performed 
better than manufacturer’s claims, but a bias in standard 
accuracy range observations was identified. 
Further lab testing indicated that, with scan averaging, 
deflections greater than 8 mm could be recovered with an RMS 
accuracy of better than ±1 mm. Shape distortion, possibly due to 
laser wavefront non-uniformity, was found and is likely the 
dominant error source precluding more accurate deformation 
estimation. Additionally, more investigation is required into the 
exact nature of the surface fitting routines that were utilised. 
Beam deflections estimated via the laser scanner measurements 
were biased, but errors were dependent upon the surface used 
for comparison. Differences between the photogrammetric and 
the top surface scanner measurements were nearly constant, 
whereas those from the bottom were not. The previously 
identified shape distortion as a function of incidence angle is 
suspected to be the dominant error source and investigations to 
quantify it are underway. 
ACKNOWLEDGEMENTS 
The authors thank Dr Ian Chandler for his lending his expertise 
in structural engineering and Terry Whiteman for his assistance 
with some of the fieldwork. 
Target 
Photogrammetric AZ 
(mm) 
Laser Scanner AZ 
from bottom (mm) 
Difference 
(mm) 
Laser Scanner AZ 
from top (mm) 
Difference 
(mm) 
1402 
-3.6 
-16.0 
12.4 
-9.6 
-6.0 
1403 
-3.7 
-15.0 
11.3 
-7.8 
-4.1 
1404 
-4.3 
-13.6 
9.3 
-10.4 
-6.1 
1405 
-5.0 
-12.4 
7.4 
-10.4 
-5.4 
1406 
-5.2 
-11.8 
6.6 
-10.8 
-5.6 
1407 
-4.7 
-11.0 
6.3 
-9.4 
-4.7 
1408 
-3.7 
-12.0 
8.3 
-7.6 
-3.9 
1409 
-3.3 
-12.6 
9.4 
-6.0 
-2.8 
RMS 
±9.1 
±4.9 
Table 3. Stringer Deflections Measured by Photogrammetry and Laser Scanning. 
-43-