Geography Reference
In-Depth Information
images displayed in real time alerts the operator to any
problems immediately and prevents a camera stopping
mid-experiment due to a full memory card. Computer
operation is particularly useful for operating multiple
cameras; images captured simultaneously can be stitched
together to capture a larger study area. The software devel-
opment kit (SDK) that is included with most cameras only
allows for one camera to be operated at a time (in other
words, a separate computer is required to operate each
camera). We are aware of only one commercially available
software that permits simultaneous operation of multiple
cameras from a single computer: Pine Tree Computing
LLC. Many camera SDKs include an option for stitching
images and batch processing. In order to successfully
stitch images there must be sufficient overlap and the
focal length of the images must be the same. It is also
important to include stitch-points, i.e., clear points that
appear in each pair of images to be stitched (assuming the
setup does not change, stitch-points are required in only
one set of images). PTGUI is a commercially available
software for stitching images that is independent of a spe-
cific camera firmware. Finally, when capturing timelapse
images it is recommended the camera be connected to an
external power source to prevent it from entering sleep
mode between shots or having a battery die mid-run.
Images typically need to be post-processed to rectify
lens distortion (curvature) and camera angle (perspec-
tive) so that distances and areas are true everywhere in the
image. The first step in correcting images is to acquire a
calibration image of the study area under the exact condi-
tions that will be used during an experiment. A calibration
image consists of a grid composed of parallel and orthog-
onal horizontal and vertical lines of known dimensions
placed parallel to the plane of the experimental surface.
The grid can be constructed, drawn onto a board, delin-
eated using stationary markers, etc. (it is recommended
to do the calibration before starting an experiment). If the
experimental setup (stream-table, camera position, focal
length) remains constant the grid can be removed and
the same calibration can be used to correct all the images.
A new calibration is required whenever any changes are
made to the experimental setup. For best results the grid
should be as close as possible in size to the field of view
of the camera. Once an image is properly corrected the
grid-lines should be straight and parallel everywhere in
the image and all distances should accurately scale with
real-world distances. A computer program can be written
to automate image correction for a series of images or
widely available commercial image correction software
can be used - Andromeda LensDoc is an example of a
commercial Photoshop plug-in that enables setting cali-
bration parameters by simply clicking on different points
in the calibration image. The parameters can then be used
to batch process a series of images. It is recommended
that all images be corrected before stitching.
Common techniques to better visualise naturally trans-
parent flow include adding coloured dye (Winterbottom
and Gilvear, 1997; Gran and Paola, 2001; Tal and Paola,
2007) and/or pigments in powder form such as titanium
dioxide (TiO
2
) to make it opaque (Martin et al., 2009).
Adding colour/opacity to the flow makes it much eas-
ier to identify both qualitatively and quantitatively from
images based on the colour value of each pixel (RGB and
HSV; colour value can be read and manipulated using
image processing software such as Photoshop and ImageJ
or using a program that treats images as matrices such
as Matlab). Jpeg image format compresses images and
therefore requires less memory than Tiff or Raw, however
the latter should be used whenever possible because the
colour value of each pixel is preserved.
A constant dye concentration can be used to estimate
the flow depth based on colour intensity (Winterbottom
and Gilvear, 1997; Gran and Paola, 2001; Tal and Paola,
2007). This technique requires uniform light-coloured
sediment; the optimal dye concentration and the vari-
ation of colour intensity with flow depth need to be
calibrated for every experimental setup. Once a mathe-
matical relationship between dye intensity and flow depth
has been obtained, flow depth can be estimated anywhere
in the image based on the pixel colour value. A simple
way to calibrate flow depth based on colour intensity is
to place tilted trays with sediment glued to the bottom
and sides and filled with dye water in the camera's field of
view. Images should be captured with a camera located
directly overhead (normal to the flow). A polariser filter
(lens) can be added to the camera in order to reduce
glare from the flow. This can be further augmented by
placing a polariser sheet in front of any lighting in order
to achieve cross-polarisation. The dye-density technique
is highly sensitive to variations in lighting which can pose
a problem in the case of large experiments. Variations
in lighting across the study reach should be identified
and multiple calibrations (i.e., trays) should be used to
minimise the error. Finally, many dyes photo-degrade
(i.e., decay) over time and need to be replenished regu-
larly to maintain a constant concentration. Once again,
multiple and continuous calibrations help account for
these changes.
Search WWH ::
Custom Search