The Use of Imagery in Laboratory Experiments - Fluvial Remote Sensing for Science and Management - page 305

grid projected onto an evolving surface (using a video or
more simply an overhead projector) and analysing the
deformation of the grid with respect to the same grid pro-
jected on a flat plane (e.g., a smooth bed at the beginning
of an experiment; Figure 13.16). The projection angle
of the grid should be sufficiently small such that bed
topography induces a significant deformation. Photos of
the plane are recorded with a digital camera positioned
vertically above the bed. The spectrum of grid patterns of
the reference image and an image containing topography
are analysed using the Fourier transform (Figure 13.17).
The local shift of the grid lines (i.e., phase-difference)
between the deformed pattern in the presence of topog-
raphy and the reference pattern is contained in the width
of the peak in the spectrum analysis and is proportional
to the local thickness of the bed (i.e. topography). The
phase difference is converted to bed elevation based on
a calibration that relates the measured phase shift to a
known change in elevation (Figure 13.17).
The simplified version of the moire method described
above has been used in laboratory physical experiments
to study granular flows (Pouliquen and Forterre, 2002),
incision dynamics of subaqueous channels (Lancien,
2005), and micro-scale braided rivers (Metivier and
Meunier, 2003). The costs involved with this type of
moire setup include a digital camera, a computer, and
a video-projector (or even less expensive - an overhead
projector), the sum of which is substantially lower than of
a laser scanner but without compromising precision. Data
acquisition and processing are relatively fast so measure-
ments can be made in quasi-real time and data processing
can be automated.
A more elaborate but more precise implementation of
the moire method uses a procedure known as phase shift-
ing to calculate the phase and robust phase unwrapping
combined with grey coding to assess the geometric param-
eters (details in Limare et al., 2011). A more sophisticated
phase-to-height conversion is based on a calibration using
a plane tilted at a known incline rather than a stationary
object (Figure 13.18). Rather than a single-grid projection,
a series of different grid patterns are projected automat-
ically in succession. Images are captured with a CDD
camera connected to the same computer generating the
grid patterns. The number of projected images can be var-
ied making it possible to switch between high resolution
and fast acquisition time according to the requirements
of the experiment.
Regardless which moire method is used, certain general
requirements must be fulfilled in order to obtain the best
raw images as input data. First, the grid projection must

1

Sparse

0.5

0

1

Dense

0.5

0

1

Very dense

0.5

0 1

414

Became wet
Remained dry

Figure 13.14 Probability by vegetation class of an area that was
vegetated at t 1 to get eroded (become wet) at t 2 versus the
probability that it will remain dry at t 2 . Probabilities were
calculated between images captured at 5-minute intervals
during all of the high flows (Tal, 2008).