Artificial Intelligence-Based Drone System for Multiclass Plant Disease Detection Using an Improved Efficient...

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Crop diseases significantly hinder agricultural development and crop quality.
• Accurate disease identification is complex due to low contrast, varying image conditions, noise, and blurriness.

Purpose of the Study:

• To develop a robust, drone-based deep learning approach for efficient and accurate crop leaf disease classification.
• To address the limitations of existing methods in handling complex image variations.

Main Methods:

• An improved EfficientNetV2-B4 deep learning architecture with added dense layers was proposed.
• The model utilizes an end-to-end training architecture for classifying deep key points of diseased crop leaves.
• Performance was evaluated using the PlantVillage Kaggle dataset and drone-captured images under diverse conditions.

Main Results:

• The proposed model achieved high performance metrics: 99.63% average precision, 99.93% average recall, and 99.99% average accuracy.
• The approach demonstrated robustness across varying image samples and capturing conditions.
• The method exhibited lower time complexity compared to existing techniques.

Conclusions:

• The developed drone-based deep learning approach offers a robust and highly accurate solution for crop leaf disease classification.
• This method effectively overcomes challenges posed by complex image variations and noise.
• The findings support the potential of advanced deep learning for enhancing agricultural monitoring and management.