Feature Mapping for Rice Leaf Defect Detection Based on a Custom Convolutional Architecture.

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Food manufacturers seek to improve quality inspection for export and safety, particularly post-COVID-19.
• Rice yield is vulnerable to plant diseases, impacting global food supply.
• Developing countries often face challenges with stringent quality inspection due to labor costs.

Purpose of the Study:

• To develop a lightweight convolutional neural network (CNN) for automated detection of rice leaf smut and rice leaf blight.
• To address data scarcity issues in agricultural applications.
• To enhance quality control in rice production and processing.

Main Methods:

• Development of a lightweight convolutional neural network architecture.
• Implementation of a Domain Feature Mapping mechanism to model data variance.
• Creation of a custom filter development mechanism with filter suppression protocols.

Main Results:

• Successful development of a CNN model for identifying specific rice plant diseases.
• Demonstrated effectiveness of Domain Feature Mapping in overcoming data limitations.
• Proposed methods provide a scalable solution for automated disease detection.

Conclusions:

• The proposed lightweight CNN offers an efficient solution for automated rice disease detection.
• The developed methods can improve quality inspection processes in rice production.
• This research contributes to enhancing food safety and export standards in agriculture.