An automated hybrid deep learning framework for paddy leaf disease identification and classification.

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Agriculture is vital in India, with crop diseases causing significant yield and quality losses.
• Rapid identification of rice diseases is critical for effective management and increased productivity.
• New technologies, particularly deep learning, offer potential solutions for agricultural challenges.

Purpose of the Study:

• To design and propose an automated deep learning model for paddy leaf disease identification and categorization.
• To enhance agricultural productivity through advanced disease detection technology.
• To improve the accuracy and efficiency of rice disease management.

Main Methods:

• Utilized a structured workflow: image acquisition, pre-processing (ROI selection, adaptive thresholding, K-means clustering), feature extraction (MobileNetV3), feature selection (GKSO-SA algorithm), and classification (CatBoost).
• Employed transfer learning with MobileNetV3 for extracting color, shape, and texture features from paddy leaf images.
• Implemented a hybrid feature selection algorithm (Genghis Khan Shark Optimization with Simulated Annealing) for optimal feature identification.

Main Results:

• The proposed deep learning model achieved a high accuracy of 98.52% in identifying and classifying paddy leaf diseases.
• The system demonstrated superior performance compared to conventional classifiers.
• Validated performance using accuracy, sensitivity, and F1-score metrics.

Conclusions:

• The developed automated deep learning model is highly effective for paddy leaf disease identification.
• This technology can significantly contribute to improving rice cultivation efficiency and yield.
• The findings support the adoption of advanced AI techniques in agriculture for disease management.