Advanced clustering and transfer learning based approach for rice leaf disease segmentation and classification.

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Accurate and early identification of rice diseases is crucial for global food security and minimizing crop losses.
• Conventional disease diagnosis methods are often inefficient, inaccurate, and time-consuming, hindering effective disease management.

Purpose of the Study:

• To develop and evaluate an improved deep learning and transfer learning framework for proficient diagnosis and categorization of rice leaf diseases.
• To overcome the limitations of traditional diagnostic techniques by leveraging advanced computational methods.

Main Methods:

• Image preprocessing including resizing and contrast enhancement using adaptive histogram equalization.
• Segmentation of diseased regions using a gravity weighted kernelised density clustering algorithm.
• Feature extraction via fine-tuning EfficientNetB0 and classification with new fully connected layers, optimized with a tent chaotic particle snow ablation optimizer.

Main Results:

• The proposed deep learning framework achieved high accuracy rates of 98.87% and 97.54% on two benchmark datasets.
• The method demonstrated superior performance and validity when compared against six other fine-tuned models.

Conclusions:

• The developed deep learning approach offers a highly accurate and efficient solution for rice leaf disease diagnosis.
• This advancement has the potential to significantly improve rice production by enabling timely and precise disease management.