hosen as
> plotted
veen the
ns of the
ov point
the laser
program
rking.
jints, 1.5
immetric
strip.
ariations
'alue has
rip on zy
uence of
1 point is
arisons.
g 7. This
but it is
does not
case, the
ia should
eviously
e it, both
ve to be
defining
us stored
different
ted with
applied.
immetric
,
3d model
"elabora-
racy val-
that the
of points
on" (62
pheres...)
in the ar-
t remains
vey.
ducation,
Research
ink: prof.
1gement,
ity; arch.
ration in
bul 2004
VISUALISATION, EXPLORATION AND CHARACTERIZATION OF VIRTUAL
COLLECTIONS
E. Paquet? and H. L. Viktor"
? Visual Information Technology, National Research Council, M50 Montreal Road, Ottawa, K1A OR6, Canada —
eric.paquet@nrc-cnrc.ca.ca
b SITE, University of Ottawa, 800 King Edward Ave., Ottawa, KIN 6N5, Canada — hlviktor@site.uottawa.ca
Commission VI, WG V/2*
KEY WORDS: Virtual Reality, Visualisation, Three-dimensional, Mobile, Data Mining, Retrieval, Query
ABSTRACT:
Unrestricted access to both historical and archaeological sites is highly desirable from both a research and a cultural perspective.
However, due to security and preservation considerations, access is becoming more and more restricted and subject to various
conditions. With the recent developments in 3D scanner technologies and photogrammetric techniques, it is now possible to acquire
and create accurate models of such sites. Through the process of virtualisation, numerous virtual collections are created that need to
be visualised, searched and eventually characterized. This paper presents a mobile virtual environment designed for the visualization
of photorealistic high-resolution virtualised scenes and artefacts. The mobile virtual environment also includes a component for
retrieving artefacts from virtual collections. This stereo virtual environment is portable and can be easily and rapidly deployed at any
suitable location, for instance an archaeological site. The architecture and the implementation of the mobile virtual environment are
described. This environment is characterized by a massively asynchronous architecture that optimises the rendering performances by
distributing the calculations over various graphical processing units. A request broker insures the synchronization among the various
components of the system. The performance of the system is illustrated through multiple examples of the visualisation of virtualised
cultural heritage sites. In addition, it is shown how it is possible to describe the geometry of the artefacts by representing them with
compact support feature vectors. A recurrent data mining system, based on these vectors, is presented. This system allows the
characterization and exploration of the collection, through cluster analysis. The system employs the “query by example” paradigm
and the knowledge of the expert in a recurrent approach, in order to identity clusters of artefacts. The virtual environment is
subsequently utilised in order to perform visual data mining on the clusters, as identified during data mining, and to characterize and
further explore the clusters by defining archetypes.
1. INTRODUCTION
The public interest toward historical, cultural and
archaeological sites and artefacts has significantly increased
over the past decade. An equivalent phenomenon has been
observed among specialists and scholars. Such phenomenon,
desirable from a cultural point of view, has introduced a new
problematic: many sites are over visited and many artefacts are
over manipulated. The consequences have been dramatic: a
gradual degradation of numerous sites has been observed and
countless artefacts have been permanently damaged or lost [1].
Several curators have reached the conclusion that the sole
solution to ensure the conservation of the artefacts is to restrict
their access to the minimum possible level. This, in turn, has
generated many questions. Who should be allowed to access
the sites? Under which conditions should this access be
granted? For instance, if the duration of the visit is too short
and if the conditions are too constraining, the visitor cannot
appreciate the site and the scholar is limited in terms of the
feasible of the research.
A proposed solution has been to virtualise such sites and
artefacts, i.c. to create an “identical” digital copy of the original
sites [2]. Initially, it seems that virtualisation is the panacea to
the above-mentioned difficulties. A virtual model provided an
unlimited access, does not suffer from degradation and can be
accessed from virtually any location. Nevertheless, one has to
recognise that such an approach conveys new challenges. For
instance, can we really certify that the virtualised site conforms
to the original? The answer is twofold. With the current
techniques, it is possible to achieve a high degree of realism.
However, it is still difficult to provide the scholar with an
accurate representation both from a chromatic and geometric
point of view. For a valuable study of this aspect, the reader is
referred to [3].
Consequently, the accuracy of vitualised sites can still be
considered as an open issue. For this reason, scholars should
have access to the raw data in addition to the final model, in
order to inspect the calibration files and to an exhaustive list of
the equipment and software utilised for the acquisition and
creation of the model. Furthermore, the scholar should be
provided with such information as the precision and the
accuracy of the model in order to be able to estimate the level of
detail to which the model can be analysed.
Despite of tremendous improvements, the acquisition and
creation of a model still involve a significant amount of manual
intervention and savoir-faire. Nevertheless, if the acquisition
pipeline is well designed [4], it is possible to acquire and create
a large amount of models. It is conceivable that in a near future,
such models will be so numerous that the creation of databases
will be justified and needed. A certain number of issues then
arise. One should be able to access the database, to search for a
