Automated Detection and Classification of Rice Crop Diseases Using Advanced Image
Processing and Machine Learning Techniques
Shashank Chaudhary
1
, Upendra Kumar
2*
1
Department of Computer Science & Engineering, Dr. Abdul Kalam Technical University, Lucknow 226031, India
2
Department of Computer Science & Engineering, Institute of Engineering and Technology, Dr. Abdul Kalam Technical
University, Lucknow 226031, India
Corresponding Author Email: ukumar@ietlucknow.ac.in
Copyright: ©2024 The authors. This article is published by IIETA and is licensed under the CC BY 4.0 license
(http://creativecommons.org/licenses/by/4.0/).
https://doi.org/10.18280/ts.410216 ABSTRACT
Received: 22 April 2023
Revised: 28 November 2023
Accepted: 20 December 2023
Available online: 30 April 2024
Agriculture, a pivotal sector in the Indian economy, plays a crucial role in national
development. A significant challenge within this domain is the detection of crop diseases,
with brown spot, leaf blast, and bacterial blight being prevalent afflictions in rice crops. This
study presents an innovative approach, integrating Gray-level Co-occurrence Matrix
(GLCM) and Intensity-Level Based Multi-Fractal Dimension (ILMFD) for feature
extraction in disease identification. The efficacy of this integrated technique was evaluated
through a comparison with various classifiers. Specifically, the Artificial Neural Network
(ANN), Support Vector Machine (SVM), and Neuro-Genetic Algorithm (Neuro-GA) were
employed to ascertain their precision in disease detection. It was observed that the
combination of GLCM and ILMFD with the Neuro-GA classifier achieved an accuracy
exceeding 90%. Remarkably, when paired with the SVM classifier, this integrated approach
yielded a precise accuracy of 96.7% in detecting brown spot disease in rice. These findings
not only validate the effectiveness of the GLCM and ILMFD methods in feature extraction
but also highlight the superior performance of the SVM classifier in crop disease detection.
This research contributes significantly to the field, offering a robust solution for accurate
disease diagnosis in rice crops, thereby aiding in the sustainable management of agricultural
practices.
Keywords:
classification, image processing, feature
extraction, machine learning
1. INTRODUCTION
India boasts some of the world's largest rice fields. It is
renowned for its extensive rice growing. Rice agriculture has
a crucial role in India's economy. Historians claim that the
cultivated variant of this plant originated in the foothills of the
Eastern Himalayas. Subsequently, it was disseminated to more
nations, including Burma, Thailand, and Vietnam [1]. Rice is
widely recognized as a highly significant staple crop. It is also
considered the second most productive cereal in the world,
following wheat. India is anticipated to have a surge in
population and a growing demand for rice, which would
consequently necessitate enhanced food security measures and
heightened agricultural output. Regrettably, illnesses have the
potential to inflict damage upon the agricultural produce of a
nation. Plant disease control is more complex due to multiple
variables. These factors encompass the rising need for secure
and nourishing sustenance, the exhaustion of ecological assets,
and the rivalry over scarce land resources. Monocultures and
agricultural intensification are responsible for the growing
threat of disease outbreaks [2]. Detecting the initial phases of
a botanical ailment is crucial to preempt its detrimental impact
on agricultural yields. Tian introduced a sophisticated
technique for identifying diseases in rice crops by utilizing a
computer vision system. The researchers employed a range of
sophisticated methodologies, including neural networks,
machine learning, and image processing, to detect and classify
plant illnesses [3]. A visual method for recognizing different
types of rice illnesses based on their distinct leaf texture
characteristics. The objective of this study is to present the
principles of image processing for the categorization of plant
diseases [4].
An inherent drawback of the existing classification systems
is their lack of generalizability. Consequently, their ability to
recognize a specific disease is limited to the extent of their
training on a particular dataset [5]. After successfully
recognizing a rice plant disease, the model had an accuracy
rate of 90%. However, subsequent exposure to varying
conditions resulted in a decline in its accuracy, so casting
doubt on its utility. This is a significant factor contributing to
the diminished precision of the outcomes. Presently, this
endeavor suffers from a want of a deep learning-based solution
capable of discerning various forms of diseases in rice plants.
The identification of diseases is a crucial aspect of
agricultural research since it enables accurate diagnosis and
monitoring of crop conditions on a farm. The objective of this
study is to enhance the efficacy of the image-based detection
system by investigating several rice leaf diseases [6]. The
primary goals are to create an automated system that use image
processing and machine learning approaches to detect and
Traitement du Signal
Vol. 41, No. 2, April, 2024, pp. 739-752
Journal homepage: http://iieta.org/journals/ts
739