Great Discoveries
1ames and Hudson,
'ramework for the
lity Approaches in
f the Internet.CAA
iffney, S. Laflin, M.
vical Reports (Int.
landscape and city
hicalco. In: The
ry, Remote Sensing
ming, China, Vol
-dimensional GIS:
poral relationships
ecta Praehistorica
02. Archaeological
1 in China. In: The
ry, Remote Sensing
ming, China, Vol
the bad, and the
ption of computer
Archaeology 1994.
n. Oxford: British
19-22.
to actuality: the
t. In: Archaeology
eilly and S.Rahtz.
umental complex of
mages. Tutorial of
logy. In: Computer
y K. Lockyear and
eports (Int. Series
'r reconstruction of
.andscape. In: The
ry, Remote Sensing
ming, China, Vol
xcavations, Maps,
4
>
Avanzini and arch.
roviding the data of
| a particular thank
Carnevali for the
THEORETICAL AND PRACTICAL ASPECTS OF ARCHAEOLOGICAL ORTHOIMAGING
D. Mavromati *, E. Petsa , G. E. Karras
* Department of Surveying, National Technical University of Athens, GR-15780 Athens, Greece (gkarras@central.ntua.gr)
? Department of Surveying, Technological Educational Institute of Athens, GR-12210 Athens, Greece (petsa@teiath. gr)
KEY WORDS: Archaeology, Orthoimage, DEM/DTM, Mosaic, Orientation, Bundle, Non-Metric, Distortion
ABSTRACT
Orthophotos are a standard requirement in archaeological documentation; yet they differ in several respects from aerial orthoimaging.
The required large scales of end-products call for close-range photography, usually taken from low altitude or with raised cameras for
horizontal recording. Special camera platforms need to be devised to this effect, such as the flexible low-cost devices (small balloon;
adapted fishing-rod) used here. With such ‘unstable’ platforms image tilt, recording distances and overlap are not easily controlled,
hence irregular strip geometries are expected. Besides, the non-metric cameras used have unknown inner orientation and often large
lens distortion. Our experiences with such bundle adjustments are discussed. Precise surface description is a further issue, more than
often involving modeling of rough surfaces with abrupt changes, discontinuities and protruding parts. Examples from different pro-
jects illustrate the authors’ experience as regards data collection allowing generation of ‘vertical triangles’, indispensable for creating
‘true orthophotos’ with commercial software. A final aspect addressed in this contribution concerns the exploitation of the numerous
existing line drawings of sites. This graphical information, mostly planar, might be extensively used as exclusive ground control to
produce orthomosaics for innumerable sites, at least as basic archival documentation. Rather than performing purely planimetric strip
adjustment, an approach is tested here which additionally makes use of suitably weighted model elevations of such planar ‘control
points" derived from the maps. The presented results show an increase in accuracy, thus indicating that in several cases existing 2D
information may help minimise, or even eliminate, the need for control surveys. The discussed aspects of archaeological ortho-
photography are illustrated with examples from various Greek sites, namely the parodoi of the ancient theatre of Sparta, the ancient
castle of Aigosthena, the ancient theatre of Zea in Piraeus and an archaic site of Zeus in Athens.
1. INTRODUCTION
Orthomosaics, based on reliable elevation information, are now
perhaps the standard photogrammetric products for archaeologi-
cal documentation. Indeed, among the deliverables asked for by
archaeological services, conventional line drawings tend to be
replaced by raster products, notably orthophotography (but also
digital developments, other projections and drapings or photo-
textured models). This is true not only for ordinary documenta-
tion but also for restoration purposes of ancient theatres, retain-
ing walls etc. Although line drawings may also be produced on
top of the orthomosaic, the latter constitutes a powerful textured
representation combining geometric accuracy with a wealth of
detail (e.g. regarding damages and decay), thus providing a suit-
able basis for conservation and restoration planning.
However, compared to conventional aerial orthoimaging, its ar-
chaeological counterpart displays a number of peculiarities. To
start with, the required large scales of the end-products call for
close-range photography. But most archaeological sites need to
be recorded either from above (as in the instance of excavations
or theatres) or using a raised camera with horizontal axis dicta-
ted by object height (as in the case of castles or retaining walls).
This poses serious questions concerning image planning and ac-
quisition. Hence, special camera platforms need to be devised to
meet the variety of requirements, as archaeological sites may be
in densely built areas or, at the other end, in isolated regions ac-
cessible only on foot. Besides, possible solutions are also limit-
ed by available financial resources which, regrettably, are often
poor. In such instances flexible low-cost devices — such as small
balloons and adapted fishing-rods — have been used for vertical
and horizontal recording (Karras et al., 1999; Petsa, 2001).
Being inherently ‘unstable’, such camera elevators involve two
major questions. First, the image tilts cannot be fully controlled.
Even if monitors adapted to the raised camera are employed (as
in one of the projects referred to later), irregular strip and block
geometries are generally expected to emerge. Imaging distances
are also not totally controllable, eventually resulting in large va-
riations in image scale. This aggravates the expected problem of
scale variations due to the often large depth extension of objects
compared to the imaging distance.
The second question is related to the fact that mainly low-cost,
small and medium format, non-metric cameras are used in most
archaeological surveys; besides, only such light-weight cameras
may be raised by the simple camera platforms mentioned above.
Of course, these cameras are characterised by unknown interior
orientation, a problem enhanced by the presence of considerable
distortion in the wide-angle lenses usually used in such cases.
The points made above underline the typical difficulties facing
phototriangulation in archaeological projects.
A further important aspect of archaeological orthoimaging con-
cerns the precise surface modeling of the site or monument, to
ensure end-products of both geometric accuracy and high visual
quality. Furthermore, accurate surface modeling is not only the
prerequisite for orthoprojection but also provides invaluable in-
formation regarding morphology and deformation, constituting
a tool in itself for the evaluation and restoration processes. In
archaeological applications, object shape is often characterised
by abrupt changes in depth and successive ‘falls’ or ‘breaks’ on
a surface which, as a rule, cannot be regarded as ‘regular’. Even
in areas which initially had a regular shape, damage often pro-
duces more complex shapes. Hence, no simple CAD modeling is
generally possible (unlike most architectural items which can be
modeled as a combination of basic regular shapes; cf. Wiede-
mann, 1996). For instance, this often entails the modeling of ‘ir-
regular’ surface patches perpendicular to the main object plane
and ‘ridges’ or strongly protruding structures. In archaeological
documentation, surface modeling and triangulation still remain
a very crucial issue (Baratin et al., 2000).
Finally, a further aspect addressed in this contribution concerns
the exploitation of pre-existing graphical plans and line draw-
ings. Indeed, numerous archaeological sites have been mapped
_413-
