bank. These small format semi-metric cameras were both
equipped with a calibrated 25 cross glass reseau plate and
coupled to enable the acquisition of synchronised stereo-pairs.
The floating marker points were constructed from black
spherical polystyrene balls. Such a design enabled the centroid
of each target to be determined using an automated digital
photogrammetric system comprised of cheap "off the shelf"
hardware and software (Chandler & Padfield, 1996). The
image coordinates of all marker points were converted into
photo-coordinates using a local bi-linear transformation
through knowledge of both measured and calibrated positions
of the reseau crosses. One of RR
the main problems which had
to be overcome was the
unique identification of point
labels for targets imaged on
each photograph of the stereo-
pair. Software was developed
to isolate the most likely
point identifier for each
measured point using
collinearity constraints. The
refined photo-coordinates
were then used to determine
the 3D coordinates of the
points using a self-calibrating
bundle adjustment and finally
coordinates were corrected by
a small vertical offset to
overcome the effects of
buoyancy.
All water surface coordinates
were imported into a terrain
modelling package to
represent the water surface.
These surfaces were then
used as: a boundary condition
for a rigid-lid flow model; and
to provide verification data
for a free-surface flow model
that calculates water surface
elevation from first
principles. The significance of
these data to such studies will
be discussed. Figure 1,
photo-control points
2. The Technique
The desire to record water surface morphology using
automated methods created several constraints which needed to
be overcome in the design of a suitable measuring system. The
water surface of a large pro-glacial stream is, of course, a
dynamic feature and the surface must be sampled at an instant
and over a sufficiently wide area required to provide adequate
input into the desired flow model. The wide daily variation of
melting snow into the pro-glacial stream causes dramatic
variations in the flow velocity. Prior experience suggested that
this velocity could be as high as 2ms', consequently any
system would have to be sufficiently robust to withstand such
velocities. The working conditions adjacent to a glacier snout
Hasselblad image showing poly-ball targets and i
(targets enhanced for reproduction)
would not be ideal, a regular power supply would be
unavailable and all equipment would have to be carried to this
remote site. The intention was to develop an automated system
of measurement capable of measuring a large number of points
at an instant. This would provide a larger sample of the surface
population than would be achievable through manual
measurement.
The unique property of the photographic image to freeze the
motion of dynamic objects suggested that a photogrammetric
solution would be most practicable. The possibility of
automating the measurement process using digital
photogrammetry also suggested potential.
2.1 Targeting
Although a turbulent water surface does create discrete natural
features which could perhaps be used to provide a natural form
of target, it was essential to obtain usable results from the field
work and so active targeting of the object was judged
necessary. Structured lighting methods routinely used for
human body-surface measurement (Mitchell, 1994) were
considered, but multiple reflections from the infinitely faceted
turbulent water surface and the distance from any power
supplies precluded such a technique. A floating target (Figure
1) was felt to be the best approach and such targets have been
used by other authors. Fraser & McGee (1995) used retro-
100
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
ref
the
de:
Col
bal
spl
tar
wa
lad
wh
On
pol
lev
the
tar
Tai
car
Im
but
pos
not
wa
wa
sys
ver
free
wel
bal
Ph
me:
the
The
effc
targ
A251
and
pur
spli
at r
was
the
0.51
plas
poi
con
Altl
pro
Syst
pro
of c
wer
inst
prol
the
3D
usir
Gec
elec
leas
pro
poir
