Developing a Classification Model to Identify Rice Plant Diseases Using Fuzzy Color and Texture Features

Rice is one of the oldest and most important cereal grains not only in India but also in the world. Roughly half of the world's population depends on rice every day. The demand for rice will exceed its production in the future as per reports. Diseases in rice plants severely affect the rice yie...

Full description

Saved in:
Bibliographic Details
Published inICFAI journal of information technology Vol. 17; no. 2; pp. 7 - 16
Main Authors Ojha, Ananta Charan, Vinitha, C
Format Journal Article
LanguageEnglish
Published Hyderabad IUP Publications 01.06.2021
Subjects
Accuracy
Agriculture
Algorithms
Classification
Color
COVID-19
Crop diseases
Crop yield
Crops
Datasets
Digital imaging
Economic impact
Feature extraction
Food
Histograms
Image processing
Machine learning
Neural networks
Pathogens
Plant diseases
Rice
Support vector machines
Texture
Asia
India
Online AccessGet full text

Cover

More Information
Summary:Rice is one of the oldest and most important cereal grains not only in India but also in the world. Roughly half of the world's population depends on rice every day. The demand for rice will exceed its production in the future as per reports. Diseases in rice plants severely affect the rice yield, thereby increasing the demand and supply gap. It results in significant agricultural and economic losses if timely actions are not taken. Identifying the diseases on time and taking the necessary measures will help reduce this loss. Digital image processing coupled with machine learning techniques can be used to identify rice plant diseases and help farmers address such issues to a great extent. The paper evaluates select machine learning algorithms in the identification and classification of rice plant diseases. The Fuzzy Color and Texture Histogram (FCTH), an unsupervised filter, has been used to extract relevant features from infected leaves of rice plants. Using a dataset from the UCI machine learning repository, experiments have been conducted to study the performance of the classifiers. The experimental results are very impressive and showed that the classification accuracies of the considered algorithms are competitive, with the Neural Network (NN) being slightly better than the rest achieving 89.583% accuracy on the test dataset.
ISSN:0973-2896