Monitoring and predicting cotton leaf diseases using deep learning approaches and mathematical models.

Area of Science:

• Agricultural Science
• Computer Science
• Formal Methods

Background:

• Cotton production is vital globally but threatened by diseases impacting yield and economies.
• Accurate monitoring and prediction of cotton crop diseases are crucial for sustainable agriculture and food security.

Purpose of the Study:

• To develop and verify a methodology for improved cotton crop disease monitoring and detection.
• To enhance the reliability of disease symptom identification using formal methods and deep learning.

Main Methods:

• Formal modeling and verification using Temporal Logic of Action (TLA+) to define and check disease symptom requirements.
• Model checking applied to TLA+ models for ensuring the correctness of disease detection properties.
• Deep learning models, specifically Convolutional Neural Network (CNN), employed for predicting various cotton diseases.

Main Results:

• The TLA+ model checking approach ensures the reliability and correctness of disease symptom detection.
• The CNN model achieved a high overall accuracy of 98.7% in predicting cotton diseases.
• High F1-scores were obtained across all disease classes, indicating robust performance (e.g., 0.90 for Powdery Mildew, 0.87 for Army Worm).

Conclusions:

• The integrated approach of formal methods and deep learning offers a reliable solution for cotton crop disease detection.
• This methodology can significantly contribute to sustainable cotton farming by enabling early and accurate disease identification.