Stacking-based and improved convolutional neural network: a new approach in rice leaf disease identification

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Rice leaf diseases significantly reduce crop yields, necessitating efficient identification methods.
• Manual disease identification is often inefficient and costly.
• Accurate disease diagnosis is crucial for timely intervention and yield improvement.

Purpose of the Study:

• To develop an efficient and accurate stacking-based ensemble learning model for rice leaf disease identification.
• To overcome the limitations of manual identification in terms of speed and cost.
• To provide a robust method for plant disease identification.

Main Methods:

• A stacking-based ensemble model was constructed using four convolutional neural networks (improved AlexNet, improved GoogLeNet, ResNet50, MobileNetV3) as base learners.
• A support vector machine (SVM) was employed as the sublearner.
• Comparative experiments were conducted using single models, different ensemble combinations, and diverse datasets.

Main Results:

• The proposed stacking-based ensemble model achieved a recognition rate of 99.69% on a rice leaf disease dataset.
• The ensemble model demonstrated superior performance compared to individual base models.
• The model also showed effectiveness on a general plant disease dataset.

Conclusions:

• The developed stacking-based ensemble model provides a highly accurate and efficient method for rice leaf disease identification.
• This approach offers a significant improvement over traditional manual methods.
• The model shows potential for broader application in plant disease identification systems.