Figure 1 : Synoptic schema of surveying process
The survey is done merging several kinds of information: bathymetry, DTM from photogrammetry, artefacts measure with photogrammetry and theoretical model of artefact objects. The entire survey is stored in a relational database and the geometry is exported toward tools for Virtual Reality (see fig. 1) . This approach will allow archaeologists to see the entire site, using immersive VR technologies, without diving. (Drap, Durand, Provin, Long. 2005).
The photogrammetric survey in Pianosa is made by a set of photographs with the right overlap (around 60%). The geometry is very similar to the technique used in aerial photogrammetry; the main difference is the distance to the seabed and the immersion in water.
As we are sure that the seabed is more or less flat, we can use a set of photographs with vertical axis to make the survey.
The photographs are taken by strips with 60% overlap for the consecutive photographs in a strip and 20% overlap from one strip to another. (See fig 3).
This first mission in Pianosa was an opportunity to test and improve several ways to perform this survey. As this site is 35m deep, we can use both a survey with divers (CNRS partner), and start a survey by ROV, managed by ISME.
The diver has a Nikon™ D70 digital camera with a 14 mm lens from Sigma™ and two flashes Subtronic™. The digital camera was embedded in a Subal™ housing with a hemispherical glass.
COMEX brought its digital camera equipped for connection to the ROV: a Nikon D2Hs, a 14 mm lens from Sigma™ and two flashes Nikon™, SB800. The housing and connector was made by COMEX with a flat glass. (See fig.2)
 |  |
Figure 2 : On the top left: the digital camera and its housing used by the diver, on the top right: the digital mounted on the housing back of the ROV, on the bottom the ROV in water with digital camera and flashes in their housing. (photo by R. Graille, CNRS
A zone to be surveyed has been determined by the team and equipped with 4 scale bar (2m) and a set of 15 makers (cement block 15x15x10cm) in order to define a network for a better ROV guidance.
In this mission ISME (LabMACS, Università Politecnica delle Marche - Ancona) has used the ROV unit Phantom S2. This ROV is an improved commercial ROV produced originally by Deep Oceans; it is a small class ROV DOE Phantom S2 with operating depth of 300 m. It is equipped with four thrusters (two horizontal main thrusters and two vertran ones) that actuate four degrees of freedom (surge, sway, heave and yaw): the onboard sensory system consists of a 3CCD camera, a deep meter, a compass and an inertial measuring unit (IMU) that evaluates linear accelerations and angular velocities along and around three axis. (Conte et Alii., 2004)
Mission tasks have also required the use and the integration in the control architecture of three sensor systems: a SONAR property of LabMACS, a rent SCOUT USBL and a Digital Photo Camera property of COMEX. The sonar heads is a MS 1000 produced by Kongsberg-Symrad and produces a pencil beam of conic shape, whose main lobe width is 2.7°. The second acoustic device used was the SCOUT USBL of Sonardyne and is equipped for ROV position tracking during the mission. Finally in order to guarantee an acquisition of high definition optical image, the COMEX camera a Nikon D2Hs with a sensor of 4.26 million total pixels was integrated in the ROV system. (Conte et Alii., 2007)
The ROV has made a survey on the zone delimited by the markers. The pilot use a video camera located on the bow. He can see the markers and pilots in order to make strips. The photographs were taken in two modes:
- Manually, an operator, looking thought the lens by a small video camera to shoot the image.
- With a fixed frequency, decided according to the ROV speed and altitude.
 |  |
Figure 3 : Two photographs from a strip made by the ROV
The camera calibration in multimedia photogrammetry is a problem already identified since almost 50 years. (Bass G., 1970) You can refer to Hans-Gerd Maas (Maas, 1995) to have an overview of the state of art of this field. The problem is not obvious, the light beam refraction through the different diopters (water, glass, air) introduces a refraction error witch is impossible to express as a function of the image plane coordinates alone. (Maas Hans-Gerd 1995)
Therefore the deviation due to refraction is close to those produced by radial distortion even if radial distortion and refraction are two physical phenomena of different nature.
For this reason we start to use standard photogrammetric calibration software and make a calibration of the set housing + digital camera. The distortion corrects in a large part the refraction perturbation. This was also shown by Kwon (Kwon, 1998) (Kwon & Lindley, 2000).
But this approach is strongly dependent of the ultimate dioptr water/glass of the housing. To try to minimize the refraction error we can found on the market some housing with a hemispherical glass, which is the case of Subal™ housing used with the diver. For the other one, made by COMEX the glass was plate and the refraction action is much more important.
We currently work on a method to compense separately refraction and distortion; this can be seen in the deliverable 1.2 and will be applied in the next mission.
For the moment and in order to validate the photogrammetric campaign, we have made the calibration using Photomodeler™ for the two housings.
The choice of a reference system to express the measured data is very important. It’s depending of the archaeological needs. Several cases can occur:
- We don’t have any way to get an absolute position, or we don’t need it. In this case we have to define the reference system on local, observable geometry. For example something which defines the axis of symmetry of the wreck (if there is one); buoys to define the vertical axis; scale bar.
- We need an absolute orientation and we have several ways to obtain it. For example a pipe line as DGPS – USBL can give an approximation of the ROV position, etc…
Figure 4 : Markers on the seabed
In Pianosa we will use an absolute reference given in two modes: when it will be possible ISME will associate for each photographs coming from the ROV six parameters as: x, y, z, Omega, Phi, Kappa. In the same time they will measure the absolute coordinates of a set of markers seen on the photographs and used as control points.
More than three hundred photographs have been taken by the diver. They cover an area of 20 x 20 meters. The orientation was done manually using Photomodeler™.
The photographs orientation was done using points on the seabed, except on the amphorae in order to be used to define a DTM on the seabed. The oriented photographs and the diver’s trajectory are visible in figure 5.
Figure 5 : Oriented photographs visualised in VRML with the non textured seabed
Five markers, visible in figure 4, were used as control points. The adaptation on these points was done outside of Photomodeler™ and the residuals are visible on the table below.
Table 1 : Residuals after adaptation photogrammetry onto acoustic survey.
(The coordinate are translated for site protection reason)
Using the oriented photographs a plotting phase, driven by archaeological knowledge is processed to obtain both 3D model representing the amphorae and a database managing all the data of the project.
- Development of the theoretical model: for each identified object, a geometrical description offers a whole of geometrical primitives, which are the only objects to be potentially measured, and a theoretical representation of the object. In our case archaeologists have identified six amphora typologyies and we shall produce a theoretical model for each of them. This theoretical model is formalized in a hybrid way, taxonomy of archaeological artefacts and an XML representation for the Amphorae typology.
- As photogrammetric measurements are highly incomplete (the object is seen only partially or may be deteriorated), an Expert System will determine the best strategy to inform provide all the geometrical parameters of the studied object, starting from the measurement process and handling the default data as defined in the archaeological model and the geometrical model. The expert System used is Jess. (http://herzberg.ca.sandia.gov/jess/)
- The resulting object is thus based on a theoretical model, dimensioned more or less partially by a photogrammetric measurement. During the exploitation of the photographs the operator can choose the number of attributes of the object which are relevant to measure. The choice of attributes will be revisable in time, as for example during a second serie of measurements. The system can be used to position in space some objects from a catalogue after a scaling process.
( http://ads.ahds.ac.uk/catalogue/archive/amphora_ahrb_2005/details.cfm?id=135 )
 |  |
Figure 6 : Direct measuring at scale 1:1 of amphorae. On the left side archeologist is measuring the amphora ; on the right a design produced at scale 1:1
At this stage the accuracy is sufficient with a measure done with two images and the interface is simpler to manage. The user has to choose two photographs, already measured amphorae are displayed and Arpenteur will generate a correspondent photogrammetric model on the fly. The application will connect to the database over the Internet to display thumbnails and to load photographs and already plotted amphorae.
In addition of the photogrammetric data all the data concerning archaeological items are stored in the database. These data are defined in the theoretical model (defined in section 3.6.1) they contain all photogrammetric data and all the archaeological data needed by archaeologists.
Figure 7 : Choosing 2 photographs from the database
and building a photogrammetric model on the fly
A direct link to the Database Php interface is available by picking the displayed amphorae in the VRML generated file.
In the case of amphorae we define four measurable zones, rims, handle, belly, bottom, and we use a set of geometrical primitives computed by least squares method onto the measured points. For example a circle on the rim or belly points, a line on bottom point and center of these two circles.
Figure 8 : Plotting amphorae according to the theoretical model
This interface (fig 8) allows the user (generally an archaeologist) to
- Recognize the amphora type on the photographs,
- Choose the amphora type in the interface combo box (The site was already studied in collaboration with archaeologists to define the typology),
- Measure a set of points on the zone where measure is allowed,
- Add archaeological comments and observations,
- Compute the object, using the measured points to construct a new instance of amphorae,
- Insure consistency between observations and theoretical model,
- Store this new instance in the remote database.
