|
You should make an attempt to capture images over the entire
exposed surface of the pile or one continuous section of the
pile with minimal overlap so that the results are not biased
by omission or by repetition. Remember that only what is visible
can be sized, and that the surface may hide variations of
the material beneath. The outside surface of a muck pile before
digging should not be used to represent the material within
the pile but may be important by itself. The surface of an
un-excavated muck pile may be quite different from the material
within the pile that is exposed while digging.
It is recommended for muck piles to let the shovel advance
to about the middle of the shot before acquiring images that
can be deemed representative of the blasted material. If you
are only able to obtain images from the exposed surface before
digging, make sure you only compare those to similar sets
of data. It is probably not good practice to compare the distribution
of the outside of an un-excavated pile to the inside of a
partially excavated pile.
The amount of fines is determined by the images at the largest
scale. The distribution of fines is calculated from the zoomed
in images of fines. Make sure that the largest scale images
include the patches of fines that are actually visible and
not just the largest boulders or your results may be in error.
You should be careful of changes in geology within the area
of interest, since most investigators are interested in the
size distribution within a geology.
Technique
Technique: To eliminate side-to-side distortion, all pictures
should be taken perpendicular to the line of the toe of the
slope. To eliminate vertical scaling error, the scaling balls
need to be placed in a manner that the balls intersect the
plane of the materials slope.
For muck piles, take 3 scales of images: 1) large scale (20
X 20 feet) including boulders and areas of fines, this scale
range is to get resolution on material above 8-inches; 2)
medium scale (10 X 10 feet) of typical regions of 2 to 8-inch
material and; 3) small scale (1.5 X 1.5 feet) are zoomed in
images of representative samples of the finer material, typically
2-inch minus. Take approximately equal numbers of images at
each scale, or more images at the large and medium scales.
If you are not interested in the size distribution of the
smallest scale of material and are happy to accept a Schuhman
or Rosin-Rammler curve in this range, you may omit taking
the small scale images.
Far-range
Mid-range
Zoom
Size Range: Make sure the largest scale images really
show the overall size range present. Include the patches of
fines that are actually visible and not just the largest boulders
or the results from the analysis may be biased toward the
coarse end.
Lighting: Be conscious that shadows and light do not
interfere with the overall image appearance. Overcast days
actually provide the best lighting due to fewer shadows. Make
sure the images are in focus.
You must get close enough so that the rock fragments are
distinguishable in the image: The image below was taken too
far away to define the particles well. Much "wasted"
space in the image, i.e. space that is not going to be analyzed
such as sky and foreground in front of the pile.
Images should contain more than a few number of particles:
Zooming: As part of your effort to capture three scales
of images, you should zoom in or get closer to patches of
fine material to help determine the size distribution of the
finer material.
Scaling: An object or objects of known size must be
in the picture in order to set the scale for the entire image
that is to be analyzed. The change in apparent size of objects
due to the top of a pile being further away from the observer
than the bottom is also corrected for using the scaling information.
One-Known Object Method: For zoomed-in images it
may be necessary to use a ruler or scale as the known size
in the image. In the case where only one known object is
used and the distance and angle are not measured, it is
important to take the image in the same plane as the slope
of the material imaged. By imaging in the same plane, scaling
differences and slope distortions are greatly reduced. The
image must also be taken as perpendicular as possible to
the scaling object, especially if the object is not spherical.
This method can also be used for images where the plane
of the material is not perpendicular to the camera. In this
case the distance to the bottom of the image must be known
as well as the angle of the pile. It is also important that
the scaling object is placed at the bottom of the image.
This makes this method a little more difficult to implement
in the field which means the two-object method is preferable
to use to correct for slopes.
Two-Known Object Method: Use two known objects to
scale the image, preferably spheres of known diameter. Place
the two scaling objects on the pile so that both are in
the field of the view of the image that is to be taken.
The objects should be at different vertical heights within
the image to correct the effect of slope on the scale. The
best scaling tools for this method are rubber balls with
handles on them so that a rope can be used to retrieve the
balls. The scaling balls need to be placed so they intersect
the plane of the slope of the material that is imaged.
Multiple Image Acquisition: A recurring and important
question is: how many images do I need to acquire? The number
of images required to calculate the size distribution of a
given sample of material is not fixed and varies from situation
to situation. The number of images to acquire depends on 1)
the physical size of population of material in question and
2) the rock size fraction that is of interest (i.e. do you
need images at all three scales to calculate a complete distribution
curve? Or, are you most interested in oversize and will accept
an estimation of finer material?). Taking these two key issues
into consideration should lead you to the correct number of
images to acquire for your sample. For more information, or
if you wish to discuss this issue, please contact Split Engineering.
Contact Split Engineering to discuss methods for determining
a significant sample for conveyor belt material.
Physical Size of Material to be Measured: If what
is on the surface of the material in question is deemed
to be representative of the entire population of material,
then images covering the entire surface area of the material
should be taken. A key consideration in assessing the surface
area is the homogeneity of the material on the surface.
If the entire pile looks similar in size on each exposed
face, then extra images of the "same" material
will probably not result in better size information. If
the surface area does expose varying size fractions, then
images of the entire surface should be acquired. Again,
when imaging muckpiles, it is recommended to acquire images
after the shovel has advanced towards the middle of the
pile as the surface of the blast is rarely representative
of the contents inside the pile.
Size Fraction of Interest: As previously recommend,
images should be acquired of the surface at three scales
(large scale - far range, medium scale - medium range and
small scale - zoom-in). This allows the software to focus
and analyze particles at different scales that will eventually
be merged together in one cumulative size distribution as
a complete size sample. For example, if you are only interested
in oversize material, you can stand back and take large
scale - far range images of the entire surface that will
capture those large particles (12" + or higher). The
Split software will not be able to detect mid-size particles
(5" - 8" or so) or smaller with only the far range
image and will estimate for what it cannot detect.
For every scale image, there is a cut-off point where the
Split imaging software cannot detect (delineate) any smaller
particles. Below that point is the "fines estimation."
Depending on the scale of the image, the fines estimation
may be for fairly large particles, not only the actual fines
in the sample. The fines estimation is based on the a Schuhmann
or Rosin-Rammler distribution and is a function of the slope
of the curve up to the cutoff point. Basically, if estimation
is not acceptable for smaller size fractions, acquire the
medium and small scale images. The Split software will merge
the entire sample together as one size distribution curve
and the size cutoff and fines estimation will be lower.
Image Acquisition Examples:
 |
Very large particles in this medium range image.
There is no need to zoom further in and acquire another
image as the particles in the image are clear and
distinguishable in the image.
If this surface area represents the entire population
of the material you wish to measure, you will have
to accept that an estimation for the particles that
you cannot see since there are no smaller particles
to measure.
|
 |
Far range picture captures largest particles,
but loses resolution on the finer particles, particularly
between the scaling balls. A medium scale image can be
acquired without moving the scaling balls to obtain better
resolution on the smaller particles.
Zoom even further in to achieve resolution on the smaller
particles to the bottom left of the top scaling ball.
If you are happy to accept an estimation of those particles
based on the slope of the curve using the large scale
and medium scale images, then omit acquiring the small
scale zoom-in image.
|
 |
How the Images Effect the Resulting Size Distribution:
Imaging Form: To keep a record of each test, the Split-Net
Digital Imaging Form (See Appendix A) should be completed
for each test analyzed. Information such as the test sample
#, photographer, sample date, sample location, sample geology,
blasting conditions, picture number, time of photo, picture
resolution, picture location, scaling method, download file
name, comments and test notes are necessary for proper image
processing and analysis. A digital imaging form should be
complete for every test imaged. An electronic copy of this
form should be included with each test sample. The more information
we have about the images, the better the results that we can
provide. This also allows us to keep records of prior analysis
and gain a better understanding of fragmentation at each site
and ultimately serve our clients better. Download a form from
the Table of Contents or contact Split Engineering and we
will e-mail a blank form to you.
|
