- 154 - 
Figure 7: The Mantova model with set of 3D edges, which were 
automatically extracted from the raw data. 
Figure 8: Top image: an orthophoto of a detail of the Genova 
model. Bottom image: the software tool, which allows 
defining the plane on which the 3D model is 
projected. 
6. CONCLUSION 
This paper presents the experiences we gathered while scanning 
and modelling different types of architectural objects. It focuses 
on the data acquisition and on the requirements for suitable 
processing software for architectural and Cultural Heritage 
applications. The findings can be summarised as follows: 
• Data acquisition: Careful and thorough data acquisition 
(for both, 3D and 2D data) is crucial for successful model 
reconstruction. Usually data processing is done off-site, so 
that missing data cannot be captured in a second pass. 
Since the ratio between acquisition and processing time is 
around 1 to 5, it is advisable to minimise the processing 
time by capturing the best data possible. 
• Data processing: Suitable software for processing the 
point clouds is a key success factors for laser range 
scanning in general. The requirements vary greatly 
between applications. Whereas for some applications it is 
sufficient to register the different scans and have a good 
management of and access to the point data, other 
applications require sophisticated modelling tools. For 
Cultural Heritage and architectural application, 
triangulation is a suitable way to model the data, especially 
when the mesh is textured with colour images. 
In general, laser range scanning should not be seen as 
competition to traditional surveying techniques, but rather as an 
addition. For simple applications traditional techniques might 
be more suitable, however in many cases laser range scanning is 
the most suited or even only way to achieve the desired result. 
The huge amount of data provided by a laser range scanner 
should not be seen as drawback of laser range scanning. The 
problem is rather the limitations of current software systems to 
handle the data. Future developments will bring further 
improvements to the processing of laser range data and thus 
make laser scanning an even more powerful tool for many 
applications. 
7. FUTURE WORK 
3DVeritas has implemented many of algorithms described in 
sections 4 and 5 and is currently implementing the parts that are 
not already included. Issues that have not been touched in this 
paper and which will become relevant for future work include 
• further automation of the algorithms, 
• closer integration of photogrammetric and scanning 
technologies, where measurements from different sources 
are combined to provide higher accuracy and 
• integration of mesh and CAD models. 
Acknowledgements 
The data used for the modelling has been provided by Prof. Ing. 
Girogio Vassena from the University of Brescia (Mantova 
model), Ing. Mario Mataloni from the University of Pescara 
(Hera model), Geom. Andrea Dessi from Cartograf Cagliari, 
(Nuraghe model), Riegl GmbH (Hamburg model) and SAT 
survey Sri (Genova model). 
References from Journals: 
F. Bemardini, J. Mittleman, H. Rushmeier, C. Silva, and G. 
Taubin. The ball-pivoting algorithm for surface 
reconstruction./CCA Transactions on Visualization and 
Computer Graphics , 5(4):349-359, Oct.-Dec. 1999. 
Besl, P. J. and McKay, N. D., 1992. A method for registration 
oO-D shapes. IEEE Trans. Pattern Analysis and Machine 
Intelligence, 14(2): 239-256. 
Canny, J., 1986. A computational approach to edge detection. 
IEEE Trans. Pattern Analysis and Machine Intelligence, 
8(6): 679-698.