tanbul 2004
Statistical
"of Applied
chnique for
dsat TM P-
& Remote
pplica-tions
on. Wuhan
niversity of
1-165.
of different
ves of the
32, pp.339-
imal bands
erpretation.
.), pp. 153-
nsor image
pplications.
23-854.
theory of
7, pp. 379-
tion from
& Remote
juantitative
ectral data
l'echnology
FROM THE ACQUISITION TO THE REPRESENTATION:
QUALITY EVALUATION OF A CLOSE RANGE MODEL
G. Tucci, V. Bonora
DINSE, Facoltà di Architettura, Politecnico di Torino, Turin, Italy - (grazia.tucci, valentina.bonora) @polito.it
F. Sacerdote, F. Costantino, D. Ostuni
Dipartimento di Ingegneria Civile, Università di Firenze, Florence, Italy - (fausto, dostuni) @dicea.unifi.it
KEY WORDS: Laser Scanning, Close Range Photogrammetry, Cultural Heritage, Comparison, 3D Modelling, Accuracy
ABSTRACT:
In this paper a multisensor approach (topography, photogrammetry, laser scanning) was exploited to generate a close range model of
a cultural heritage object in order to evaluate the accuracy of data in all steps of the model production, from the acquisition to the
representation.
A programmed data redundancy allowed to verify the accuracy of each technique as well as the presence of possible surveying errors.
At the end, a comparison from data acquired by different tecniques was done in order to verify the accuracy of the laser model. We
have explored the entire workflow to obtain a complete 3d model: range image registration, points decimation, triangulation, mesh
editing, model texturing. The results are here presented.
1. INTRODUCTION
New three-dimensional models can be useful for programs of
analysis and conservation of cultural heritage. Models make
easier the understanding and the communication of particularly
shaped architectonic structures, through a synthetic vision. A
multisensor approach seems to be the most suitable solution for
articulated geometry. Our study case is the transept-apse complex
of S. Francesco al Prato in Perugia, one of the most important
Franciscan Friars churches, the second after that of Assisi.
To acquire metric data, different procedures have been employed:
topography was used to establish a reference system and to
acquire natural and target control points; photogrammetry was
employed to obtain breaklines and sections from an accurate
restitution, in order to compare them with laser scanning data;
and laser scanning. The model has been obtained from the merge
of these different acquisitions in order to evaluate the accuracy
of data in all steps of the model production, from the acquisition
to the representation. A programmed data redundancy allowed
to verify the accuracy of each technique as well as the presence
of possible surveying errors. At the end, a comparison from data
acquired by different tecniques was done in order to verify the
accuracy of the laser scanner model.
2. DATA ACQUISITION: METHODS AND
INSTRUMENTATION
The articulated geometry, derived from a mixture of architectonic
elements, structural damage and superficial decay, required the
survey of many points, imposing the management of numerous
detail drafts to choose the representative elements.
Topographic data: about one thousand detail points were collected
with a reflector-less total station Leica TCR703, after having traced
a small net of arrangement; the adjustment were performed by
using the least squares method. At the same time, about 100 con-
trol points were measured - natural points on the transept, 40 tar-
gets on the apse and 17 specific Cyrax targets on the entire scene
- to orient photogrammetric stereoimages and to define a refer-
ence system for range images.
Photogrammetric data: the images were acquired by means of
both semimetric Rollei 6006 camera (f=40 mm), and digital Nikon
D! camera (f=24 mm). With semimetric camera, three
stereocouples were acquired on the apse, with mobile scaffold-
ing, obtaining about 1:350 scale of images. On the transept, twelve
stereocouples were acquired at about 1:250 scale (longitudinal
overlap =80%, transversal overlap = 40%). The images were ori-
ented and restituted with both analytic stereoplotter Digicart 40
(by Siscam) and digital Stereoview (by Menci Software).
Laser scanning data: for the transept, Cyrax 2500 laser scanner
(by Leica) were used, with sampling step of about 1.5 cm; in this
way all of the transept was covered. For more complex details,
as for example, the capitals, a sampling step of 6 mm was set up.
In all, 20 range maps, with an overlapping of about 40%, were
generated. This overlap has been performed not only to align 3D
images but also to cover undercuts and hidden zones. The laser
scanner were mounted on its tripod or simply placed on the scaf-
foldings, at various levels, to avoid great inclinations.
3. DATA PROCESSING
Increasing automation of data acquisition and processing is
essential to widen the use of the 3D model, but despite this, at
present, to represent and to elaborate data, is still, unavoidably, a
selective operation: what changes is the moment in which synthetic
interpolation, to obtain structured data, is required. With every
survey technique we obtain a synthetic model of reality, but
different cognitive approaches are available derived from different
measurement techniques: topography and photogrammetry data
collection phase requires a preliminary interpretation process,
while laser scanner acquisition collects redundant data, and
