Environmental Engineering Reference
In-Depth Information
illumination source and target area. This assembly
helps darken bright areas created by translucent par-
ticles and reduces scattered light caused by refraction
and refl ectance of the material being imaged. A suit-
able diffuser is needed for the backlighting to assure
balanced lighting throughout the image.
Turbidity, caused by organic and colloidal mate-
rial, is another hindering factor in obtaining an
assessable particle image for analysis. The use of a
near-ultraviolet wavelength of 0.45-0.5
particles that appear as a single blob on the image.
The software's segmentation algorithm works well
in identifying discrete particles within aggregates by
detecting disparities within clusters. Because of this
and other possible hindering factors, it is desirable
to analyze several images from the same water-sed-
iment sample to better characterize the actual volume
SSC computed from poor-quality images. The soft-
ware is designed to analyze selected layers of the
image starting with well-delineated and easily iden-
tifi able particles, leaving characterization of those
particles that are obscured or that otherwise present
defi nitional problems for the fi nal and most compu-
tationally intensive analyses. Research on the photo-
imaging technology continues to focus on refi ning
the software to maximize automatic interpretation of
aggregates.
For example, the software is able to distinguish a
blob as two discrete particles, labeled as 100 and 102
(numbers appear above respective blobs) (Fig. 1.15).
Although the blob labeled as 99 may be two con-
nected or overlapping particles, the software inter-
preted the blob as a single particle. Very fi ne sand
composes the sample material used in this image.
Using a microscope, it was observed that some of the
sand grains are indeed made up of two naturally
fused minerals that gave some of the single particles
a barbell-shape appearance.
m produces
a sharper image of particles. Also, with the shorter
wavelength, there is less light scatter due to refl ect-
ance and refraction as occurs when using the full
visible light spectrum. Figure 1.14 shows suspended
material fi ner than 62
μ
m at an SSC of 10 g/L
(10,000 mg/L) in a sample that was seeded with a
small number of 125- to 250-
μ
m particles that were
digitally enhanced by the software.
In some cases, the binary image could still be
degraded by turbidity, depending on the nature of
the factors causing the turbidity. If the spatial cor-
relation of the background cannot be automatically
resolved, automatic detection of particle boundaries
becomes less precise or unattainable. More analysis
and development is required in this regard.
Perhaps the most diffi cult task in the automatic
calculation of size characteristics of imaged blobs
deals with connected, aggregated, and overlapping
μ
Fig. 1.14
A morphologically transformed image of a
water-sediment mixture composed of 10 g/L of material fi ner
than 62
Fig. 1.15
A morphologically transformed image of a
water-sediment mixture composed of 62-125
m particles
showing potentially inconsistent interpretation of overlapping
or connected particles.
μ
μ
m, seeded with 125- to 250-
μ
m particles that appear
as dark blobs.
