Image Detection and Classification of Newcastle and Avian Flu Diseases Infected Poultry Using Machine Learning Techniques

Abstract

The frequency at which diseases occur in poultry nowadays is staggering. Poultry diseases, such as Newcastle
disease, Avian Influenza etc. usually brings about serious economic losses to poultry business owners and also to
farm produce consumers. Prompt warning and identification of emerging poultry disease outbreaks is of utmost
importance in the poultry business. Digital imaging technology and machine learning algorithms have made room
for the effective observation / monitoring of poultry health status via surveillance cameras online and in real time
has proven to be an effective way to prevent large-scale outbreaks of diseases. To analyze the images of healthy
and diseased birds, images of healthy birds were taken directly from poultry farms using different camera devices
such as Digital cameras, Mobile Phones etc. The first step we took was to transform the images into a fixed sized
length of dimension (64, 64, 3). The images were then augmented. Firstly, to help increase the size of the dataset,
Secondly, to create variations that will better capture reality, so as to increase the ability of the model to generalize
better and predict out of sample data more accurately. Other augmentation carried out on the training set include
Scaling, Rotation, Shifting, Zooming and Flipping (horizontally). Using the Models implemented in this research,
accuracy rates of 95% and 98% are obtained. The results show that digital image processing and the machine
learning algorithm implemented in this research can effectively detect and classify sick/diseased birds from
healthy Birds whilst giving high accuracy and good performance which will aid in giving early warning signals.
This research can serve as a reference for the intelligent detection and classification / identification of sick birds
from healthy birds.