Abstract :
Abstract — A study is done to develop fluvial particle
monitoring system based on application of spectral imaging,
image processing and artificial neural network. Research in
this field has been initiated from 2004 at Kathmandu
University.
This research is applied to Bagmati river which is the
principal river flowing and draining across Kathmandu
Valley. Fluvial particles of Bagmati River were continuously
monitored at 18 different strategic locations. These samples
were continuously monitored in the Lab of Machine Vision at
Kathmandu University in separate pre-monsoon, monsoon
and post-monsoon seasons of 2011. A lab set up equipped
with image processing and spectral imaging is developed to
monitor the contents of fluvial particles. The set up resembles
as a river path that contains upper reservoir and lower
reservoir and a special transparent flow cell fabricated in
between to monitor the particles instantaneously. An
application is developed to characterize the particles that run
on computer using Matrox imaging Library software and
Mat lab 6.5 environments.
The samples were taken in standard 125 millilitre jar. It
characterizes organic and inorganic according to the spectral
signature of the particles trained with Artificial Neural
Network. Organic particles have different signature than that
of inorganic and are most distinct at 630 and 670nm
wavelength. Based on these characterizing properties
samples from 18 strategic locations of Baghmati rivers were
characterized. Reflectance property is used to characterize
particles with Perceptron neural network and hardlim as a
transfer function. The result depicted that the ratio of
organic to inorganic is found to be 0.1111 at upstream of
Sundarijal, 0.13889 at Guyeswori and 0.16 at Chobhar
according to weight basis. Chobhar is the spot with high
amount of both organic and in-organic contents. According
to particles count the ratio of organic to inorganic is found to
be 0.02362 at upstream of Sundarijal, 0.02955 at Guyeswori
and 0.034576 at Chobhar. So this result shows that this
system can be applied for water particles monitoring. This
system can be used to find out the socio-economic activities at
different points, suitable waste-water treatment plant area,
irrigation and aquatic life preservation.
