) 1,60 meters and an 
have been supposed, 
lia wrecks specimen 
ictory stamps haven't 
s of the clay refer to 
ania and Latium. On 
s are visible, used as 
or repair these large 
ic amphorae small 
1 the site too. Only a 
finitive results: from 
ng to the kind of the 
with very big dolia, 
ssible to know if the 
wards the provinces 
roupe wrecks) or the 
yusse, Diano Marina 
th dolia belong to a 
ntury B. C. and the 
Dressel 20 fragment, 
the date around the 
| by eleven blocks of 
1 drum rough-hewed 
neters from the dolia 
oral reef, at a ca. 4-5 
the original stowage 
e ship. We can't see 
? have been crushed 
| by a layer of sea 
unauthorized fishers 
retions and chips do 
ments and to read 
y. 
for all the foundings 
v that these specimen 
iarble caves (bianco 
ninate the original 
olute certainty. The 
| 9 tons ( block n. 3) 
damaged amphorae 
] of I century B. C.- 
only clue to date the 
Id have been carried 
| by Mistral, West, 
vas sailing along the 
Mountains marbles 
the middle of the I 
age. If it is a Roman 
ge ship, because we 
ads from 100 to 200 
id it connected Luni, 
rble, to Rome, where 
y were carried along 
tisfy the continuous 
te architecture. 
1.2.4 Traditional survey 
The traditional graphic survey of the two sites, particularly the 
second, has been made in several following dives, by means of 
underwater measurements, hard and tiring (subjected to human 
error), from a point zero. The photogrammetry and three- 
dimensional modelling program on the two sites has been 
started in 2001 by Soprintendenza per i Beni Archeologici della 
Toscana (Pamela Gambogi) in cooperation with Marco 
Canciani from University of Roma Tre and Pierre Drap from 
CNRS in Marseille (France). The purpose is to obtain models 
much better and more precise than the traditional relief and 
rarely used for underwater sites. 
2. PHOTOGRAMMETRIC APPROACH 
Underwater surveying is a non-obvious work, not very accurate 
and requires a lot of time which is not very convenient in an 
underwater context. 
Thanks to a new ‘CIPA, Heritage, Documentation’ task group, 
dedicated to underwater photogrammetry and created in 
September 2001, we take advantage of this multidisciplinary 
team, represented by the diversity of the co-authors. [Cipa-uwp, 
2002]. The general way to do was inspired from the experience 
on the Etruscan wreck surveyed in France, ([Drap, Long, 
Durand, Grussenmeyer, 2001-A], [Drap, Bruno, Long, Durand, 
Grussenmeyer, 2002]) both for the survey phase and the data 
management methodology. 
  
Figure 2. Photograph for the marble survey. On the left side a 
buoy with two perforated table tennis balls, on the 
right side a ruler for scaling. 
2.1 General method 
The photogrammetric campaign was done on July 2001 near 
Livorno, Italy with the Italian team of Firenze, (Soprintendenza 
per i Beni Archeologici della Toscana) Pamela Gambogi, 
director of the excavation, in addition we have to thank Rolland 
Graille, professional diver from the CNRS “Centre 
d'oceanologie de Marseille" who came to help us with his 
competence and material. 
We decided to use a very simple and cheap way : digital Nikon 
Coolpix 990 in a Ikelite housing, *aerial photography" type for 
the survey, and Photomodeler software for the orientation and 
plotting phase. 
2.2 The survey 
2.2.1 Shooting photographs 
The survey phase was done as an aerial photogrammetry but 
without any control points. 
Curiously one of the main difficulty was that the piece of 
marble which was shallow. The depth was approximately three 
meters and the photographer-diver was near the surface which 
  
Figure 3. Refracted and non refracted coordinate of measured 
point on three square rims [Kwon, 1998] 
is really problematic for the stability. In underwater 
photography it's impossible to be far from the object (due to a 
lot of suspended particles) . 
2.2.2 Orientation 
The selected method was to minimize the time of intervention. 
A set of rulers to put the model in scale and several buoys for 
vertical reference. (The buoys were made with some perforated 
table tennis balls). 
2.3 The calibration process 
The camera calibration in multimedia photogrammetry is a 
problem already identified since almost 50 years ([ASP, 1980] 
p. 838). You can refer to [Maas, 2000] to have an overview of 
the state of art of this field. The problem is not obvious, the 
light beam refraction through the different dioptre (water, glass, 
air) introduce a refraction error which is impossible to express 
as a function of the image plane coordinate alone. 
A lot of authors give some solution more or less simple, usually 
with iterative process, in order to solve this problem. [Kwon, 
Lindley, 2000.] [Maas, 2000]. Even if Hans-Gerd Mass give a 
solution simplified it refers always to a ‘standard case’ of 
multimedia photogrammetry : only three media, an object in 
liquid, a plan-parallel transparent plane making the separation 
between the object and the sensor located in air. 
[Kwon, 1998] gives a representation of this distortion as a pin- 
cushion distortion in a particular case : a set of control points 
are marked on a plane, the control plane is parallel to the image 
plane and the camera axis passes through the center of the 
control plane. In these condition we have a distortion 
represented in Figure 3. 
To solve this problem in standard condition (normal camera 
   
Figure 4. Photograph for calibration process with Photomo- 
deler. 
—329—