Geography Reference
In-Depth Information
Step 1
Step 2
Step 3
Collection of photos
Extraction of grains
Analysis of results
What sampling area
is required?
Does the method of
calculating percentiles
affect the results?
What is the size of the minimum
grain identifiable?
[ Graham et al ., 2005a]
What internal camera
settings (focal length,
resolution etc.) are required?
[ Graham et al ., 2005a]
Consequently:
What effect does truncation
have on the resulting size
distribution?
What external camera
configuration (elevation,
flash etc.) is required?
[ Graham et al ., 2005b]
Does bed structure affect the
grain-size distribution?
What type of illumination is
required?
[ Graham et al ., 2005b]
What image resolution is
required?
[ Graham et al ., 2005a]
How should particles touching
the edges of the photo be treated?
[ Graham et al ., 2005b]
Figure 15.4 The three steps in deriving grain-size information from ground-based digital photographs, illustrating some key
procedural questions associated with each. From Graham et al. (2010) Copyright 2010 American Geophysical Union.
the image and the image scale to determine their size
(Figure 15.4). Such approaches are a development of
earlier, manual, methods of analysis using analogue pho-
tographs (e.g. Adams, 1979; Ibbeken and Schleyer, 1986;
Church et al., 1987; Diepenbroek et al., 1992; Diepen-
broek and De Jong, 1994; Ibbeken et al., 1998). The
development of automatic algorithms for extracting grain
boundaries from digital images of in situ sediments is a
comparatively complex problem because of the between-
and within-site heterogeneity (e.g. of size, packing con-
figuration, shape, surface texture and colour) of natural
sediments and the difficulty of controlling environmen-
tal conditions (e.g. ambient lighting levels and light
direction). Nevertheless, considerable success has been
achieved, with errors comparable in magnitude to those
associated with traditional field-based grid samples (But-
ler et al., 2001; Sime and Ferguson 2003; Graham et al.,
2005a & b; Strom et al., 2010) whilst field time is reduced
to a few minutes per sample. The key advantage of these
methods over empirical techniques is that - because they
measure each grain directly - they can provide complete
grain-size distributions. They have also been shown to
be transferable between sites with different petrography,
grain size and sorting, and packing configuration without
the need for recalibration (Graham et al., 2005b).
Key limitations of object-based approaches are that
lighting conditions need greater control than for empirical
approaches (to maximise contrast; Graham et al., 2005a)
and there is a minimum resolvable grain size that varies
as a function of image resolution (resulting in truncation
of the fine end of the grain-size distribution). Lighting is
relatively easily controlled by shading the sediment from
direct sunlight and using a camera-mounted flash. The
truncation of small grains will become less significant as
camera technology develops, but the effect on coarser
percentiles (above D 50 ) is minimal except where the sedi-
ment contains more than 5% sand (Graham et al., 2010).
It might be expected that sediment structure would have
a significant effect on photographically-derived grain-size
distributions (as a result of partial burial of grains, over-
lapping grains, and foreshortening). The magnitude of
such effects is difficult to determine, but Graham et al.
(2010) concluded that they are small relative to other
sources of error.
One further issue is to ensure that photographs are
sufficiently large to include a representative grain-size
sample. Grain-size percentiles have been found to stabilise
 
Search WWH ::




Custom Search