Abstract:
With growing demand of food, the world has adapted synthetic fertilizers so rapidly. Due to
which there are other critical issues rising every day including the fact that food we eat is not
much healthy. Also, synthetic fertilizers are increasing pollution level in the world. There is a
need to create awareness for natural ways of growing plants rapidly.
Microbes play an important role in maintaining a natural environment for plant growth and
help plant to resist for diseases naturally. Some of these microbes are responsible for nitrogen
fixations. Some of them help absorbing phosphate. Hence perform many other roles. They can
also be pathogenic or nonpathogenic. A farmer need to know about the microbes present in the
soil so that he can know that the soil is good for which kind of plants or how can he treat his
soil so that useful microbes are not affected.
We all know that automation is the future. The purpose of this research work is to get an
automated solution to facilitate the farmers keeping in mind the problems mentioned above.
We are training a classifier for identified plant growth promoting bacteria. So that there will be
no need of a long process of identification for a Bactria obtained from soil sample. Just once
bacteria are identified and DNA has been found, classifier is trained for their dataset. Every
next time it will be able to identify that bacteria.
A data set has been generated by microscopy for Plant Growth Promoting Bacteria (PGPB), at
100x. Images are taken from several different slides. Covering various areas of a single slide
without repeating an area. Different classifiers including alex Net, VGG16, Google Net have
been trained for the mentioned dataset to compare the accuracy and results. The layers of the
features have also been variated for various experiments to see the effect on accuracy. MATLab
is the platform used for the experiments.
