EXPERIENCES IN PHOTOGRAMMETRIC ARCHAEOLOGICAL RECORDING 
D. Mavromati a , E. Petsa b , G. E. Karras a 
a Department of Surveying, National Technical University of Athens, GR-15780 Athens, Greece (gkarras@central.ntua.gr) 
b Department of Surveying, Technological Educational Institute of Athens, GR-12210 Athens, Greece (petsa@teiath.gr) 
KEY WORDS: Archaeology, DEM/DTM, Laser scanning, Orthoimage, Bundle, Non-Metric 
ABSTRACT 
Photogrammetric textured representations of archaeological sites, and above all orthorectification, combine geometric accuracy with 
visual detail (regarding damage or decay), thus providing a suitable basis for conservation. Yet, orthoprojection in archaeology still 
poses difficult problems. At the examples of two ancient Greek sites, certain issues are discussed and illustrated. First, in most cases 
sites need to be recorded from considerable heights above ground with special low-cost camera platforms (balloon, modified fishing- 
rod etc.). A usual consequence of ‘unstable’ camera elevators is poor control over image rotations, responsible for irregular strip geo 
metry; bundle adjustment is further complicated by unknown interior orientation of lightweight non-metric cameras and strong dis 
tortions of wide-angle lenses. A second crucial aspect dealt with here is the authors’ approach for precise surface modeling to ensure 
products of both geometric accuracy and high iconic quality; this entails surface description through a careful combination of break 
lines and densely sampled spot elevations for handling edges and surface discontinuities. Regarding laser scanning, now being exten 
sively tried in the context of archaeology, experiments carried out here confirmed that it could indeed replace tedious photogramme 
tric 3D modeling in several cases. However, it is rather clear that laser scanning cannot in fact totally replace photogrammetric mo 
deling. This is due not only to problems posed by shape, size, location and surroundings of many archaeological objects, but also to 
problems emerging mainly with respect to edges. It is concluded that simple means of image acquisition and careful photogrammetric 
handling can produce results of high geometric and visual quality, while tiresome photogrammetric modeling can partly (but some 
times cannot) be replaced by laser scanning. Functional synergy of the two approaches is a delicate matter to be further investigated. 
1. INTRODUCTION 
Among deliverables required by most archaeological services, 
line drawings now face an ever-growing competition from raster 
products, notably orthomosaics, which are now a standard pho 
togrammetric product. Indeed, these constitute powerful textur 
ed representations, combining geometric accuracy with a wealth 
of detail, suitable for conservation and restoration planning. But 
archaeological orthoimaging has its own peculiar aspects when 
compared to its conventional aerial counterpart, as recently out 
lined by Mavromati et al. (2002) based on experiences from the 
collaboration with the Greek Ministry of Culture. 
Summarily put, archaeological sites generally need to be record 
ed from above or horizontally using a raised camera. According 
to location - densely built areas, sites accessible only on foot - 
and often under constraints of poor financial resources, flexible 
low-cost camera platforms must be devised to meet a variety of 
conditions (Karras et al., 1999; Mavromati et al., 2002). Even if 
monitors are adapted to cameras, however, such inherently ‘un 
stable’ elevators do not allow full control over image tilts, being 
thus responsible for irregular strip/block geometry. The same is 
true for imaging distances and resulting scale variations, further 
aggravated by large object extensions in depth (these also cause 
strong differences in perspective between adjacent images). Fur 
thermore, mainly ordinary light non-metric cameras, of small or 
medium format, are used in this context. The unknown interior 
orientation and considerable distortion of wide-angle lenses, 
usually used in these cases, add to the typical difficulties facing 
phototriangulation tasks in archaeology. At least to the authors’ 
experience, archaeological recording under such circumstances 
appears as a rather ‘generic’ photogrammetric problem. 
Not least, of course, among the peculiarities of archaeological 
recording one meets precise 3D surface modeling to ensure geo 
metric accuracy and visual quality of results. Not only a prere 
quisite for orthoprojection, accurate 3D modeling also provides 
useful information on morphology or deformation, being a tool 
in its own right in evaluation and restoration processes. Unlike 
several architectural items, shape of archaeological objects may 
often be very irregular (one is tempted to say ‘arbitrary’), distin 
guished by extreme changes in relief, ‘breaks’, ‘ridges’, edges or 
discontinuities, a substantial part of which have been caused by 
damage. This entails modeling of surface patches almost per 
pendicular to each other or strongly protruding structures. Sur 
face triangulation, under these circumstances, is a highly crucial 
issue. The authors discuss and illustrate their photogrammetric 
approach in this respect (see also Mavromati et al., 2002). 
Of course, photogrammetric surface point collection is mostly a 
tedious manual task, as automatic DSM generation in archaeolo 
gy still remains an open question (Baratin et al., 2000). Laser 
scanning, on the other hand, is a powerful technology, capable 
of collecting vast numbers of surface points in far shorter times, 
and can thus provide the 3D support for orthoprojection (Monti 
et al., 2002). However, apart from the high cost and the obvious 
difficulties (large volume of data, difficult to manage; noise), a 
laser scanning approach faces problems of post-processing for 
creating triangulated meshes suitable for the existing orthophoto 
software (Bohler et al., 2001; Balletti & Guerra, 2002); perhaps 
the question of surface discontinuities is, among these, the most 
important (e.g. Boccardo & Comoglio, 2000). And besides, not 
every archaeological site is accessible to laser scanners as it may 
be to photography. 
The aspects of archaeological orthoimaging, referred to above, 
are discussed and illustrated at the examples of two Greek sites. 
2. BUNDLE ADJUSTMENT 
As also outlined in Mavromati et al. (2002), this is a key issue 
in archaeological orthoimaging, irrespective of the mode of 3D 
modeling. For reasons mentioned above, recovering reliable ex 
terior orientation parameters may well not be a trivial task. Re 
garding interior orientation, full self-calibrating bundle adjust-