Abstract:
Classification of insect species in agricultural sector is still very challenging even though
a lot of work have been done in this area. The tininess of pest or insect and the difference
of their structure make this job even more difficult. Nowadays almost every classification
is based on manually engineered features like scale-invariant feature transform (SIFT)
and the histogram of oriented gradients (HOG). But here, we will use Convolutional
Neutral Network (CNN) which is a feature learning algorithm. Here the task is to train a
pretrained model to detect and drag out necessary feature needed for the particular task. It
detect various pest in the crops like paddy, wheat, corn, pulses, potato, sugarcane etc. We
want to develop a pest detection architecture which is based on image processing and for
this approach we are using huge amount of pest image and all of the pest images are
learned from a huge amount unlabeled image patches using of supervised learning
methods and Convolutional Neural Network (CNN) algorithm. Throughout the study we
will be using some captured previously captured pest image with different background
then try to identify the presence of pest in the crop by processing the data using deep
learning algorithm. A sample dataset of pests will be using for this work by deep learning
algorithm. We will use some preprocessed model of tensorflow. We think we will be able
to develop a system that will give us more accurate results in pest detection, because this
system is able to detect different types of pests easily. We will try to reduce crop damage
using our system.