A novel ensemble learning method for crop leaf disease recognition.

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Deep learning models are crucial for crop disease recognition.
• Diverse crop types and diseases present challenges for model generalization.
• A unified model for optimal recognition across all crops/diseases is difficult to achieve.

Purpose of the Study:

• To propose a novel ensemble learning method for crop leaf disease recognition (ELCDR).
• To address the generalization challenges of deep learning models in crop disease identification.
• To improve recognition accuracy and performance metrics for various crop diseases.

Main Methods:

• Developed ELCDR, an ensemble learning method for crop disease recognition.
• ELCDR assigns weights to models based on their feature extraction performance, measured by feature vector distribution.
• Evaluated ELCDR on disease images from four different crops.

Main Results:

• ELCDR significantly improved accuracy compared to optimal single models across apple, corn, grape, and rice.
• ELCDR demonstrated superior performance over traditional voting-based ensemble methods.
• Improvements were also observed in precision, recall, and F1-score metrics.

Conclusions:

• ELCDR effectively enhances crop leaf disease recognition accuracy and performance.
• The proposed weighting strategy based on feature extraction performance is a key innovation.
• ELCDR offers a promising solution for generalized crop disease identification.