DNS
AM aspects
er ? and into
try
e put on the
ju dans une
hnique de la
ss ci-dessus.
or pits for
st purposes,
Ty is ready :
vhich allows
nt (with or
e the results
ies the study
d» and will
er and the
ture,
ft» and will
ta.
neasurement
factor that
successful
ist ten years,
using the technique of film-based photogrammetry but
big restraints remained, mainly :
e validation of data capture only at the end,
e longtime for data processing.
CCD sensors, once the reachable accuracy has been
sufficient, appeared as the solution: the user can
immediately check the images and can get the results
very quickly.
CCD sensors available were at first such cameras as the
KODAK MEGAPLUS with 1.4 millions pixels and now
42 and even 6.3. Practically, this means that the
resolution of the sensor is an old problem and future
improvements will be more pertinent on other sides.
Canisters are available today especially because of the
development of vision applications. The only problem is
the optical quality of the glass and the isolation. Good
results can be reached with an optical quality of A/4 and
the isolation can be solved for example with a water-
cooling system for a thermal consideration or a lead
protection for a nuclear consideration.
2.2 Solutions for methodology
The range of applications that ESIC has recently solved
leads to distinguish two types of applications :
e concerning the mutual positioning sensor/object :
object fixed and sensor moving or object moving and
sensor fixed,
e concerning data capture and processing:
stereoscopic configuration — or monoscopic
configuration.
Concerning the mutual positioning sensor/object, both
cases (object fixed and sensor moving or object moving
and sensor fixed) have advantages and drawbacks. From
our experience, the most accurate is the first one, object
fixed and sensor moving because :
e there is no possible object deformation,
e there is a big degree of freedom to optimize the
geometrical configuration.
On the other hand, it is more complicate to move the
camera than to put the object on a rotating table.
So each case has to be thought knowing that the required
accuracy will mainly decide.
The other aspect of the methodology is the distinction
between the stereoscopic and the monoscopic
configuration. Here the mainly criteria are :
e the required accuracy (stereoscopic configuration
being less accurate),
e the possibility of putting targets or not on the object.
The possibility of putting targets or not has to be
considered more in detail. In fact, the stereoscopic
approach had been used many times and had given quite
good results. But we have now to distinguish the
surfaces on which no «natural point» can be well defined
and the surfaces on which some «natural points» as holes
or edges of planes can be well identified ; in such a case,
the convergent approach can still be used with the only
effect of loosing accuracy.
131
2.3 Solutions for soft
Solutions for soft are concentrated on the influence of
two components :
® the separating glass,
e the object medium.
Here the purpose is to reach the sufficient accuracy,
knowing that, in case of bad accuracy required, many
things can be neglected. Good accuracy today, while
waiting for new developments, can be reached with
some artifice ; the most interesting one is to put a known
object into the scene which allows to modelize the
corrections for all the xy measured coordinates on the
image.
3. RADIOACTIVITY
In a nuclear environment, requirements turn mainly
around :
e interchangeability,
e as-built model for maintenance or dismantling.
ESIC carried out its first measurement in such an area in
1992 with the prototype of the V-STARS (GSI:
Geodetic Services Incorporation) system (see figure 1).
The sensor used at that time was a KODAK
MEGAPLUS camera of 1.4 millions pixels and the aim
of the measurement was to define the positions of
interfaces of an equipment that has to be put into a
radioactive cell.
The limits found at that time were :
e in terms of radioactivity : 10^ rad/hour,
e in terms of accuracy : + 0,2 mm
(for a volume of 2m x 2m x 2m)
Such results opened the way for accurate measurements
in radioactive areas.
One particularity that has to be mentioned is that, to
decrease the level of radioactivity for the sensor, a
mirror (45?) was implemented in front of the camera.
Final results of these kinds of measurements, depending
on the way that images are processed, can be : either a
numerical 3D analysis for interchangeability, or a
complete 3D as-built model (see figure 1.2) with as-built
drawings, or drawings for manufacturing.
4. UNDERWATER
When the requirements, in terms of accuracy, are very
high, underwater measurement seems to be the most
complicated job. Water has to be clean and stable,
lightening requires good experience and data have to be
processed taking into account the optical deviations. But,
once all this is done, many measurements are possible.
One example carried out by ESIC is concerning the
measurement of rails and of a carrier interface (see
figure 2.1). This measurement is usually done by divers
with all the run risks of such situations ! (and the
poorness of the information !).
For this measurement, the required accuracy was + 5
mm and the reached accuracy was + 3 mm.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
