A Diffusion-Based Detection Model for Accurate Soybean Disease Identification in Smart Agricultural Environments.

Area of Science:

• Agricultural Science
• Computer Vision
• Artificial Intelligence

Background:

• Traditional soybean disease detection methods struggle in complex field conditions.
• Intelligent agricultural management requires accurate and robust disease identification.

Purpose of the Study:

• To develop an advanced diffusion-based object detection model for enhanced soybean disease detection.
• To improve performance in complex backgrounds and diverse disease regions.

Main Methods:

• Proposed a diffusion-based object detection model incorporating an endogenous diffusion sub-network and loss function.
• Optimized feature distributions for progressive enhancement.
• Introduced an endogenous diffusion attention mechanism.

Main Results:

• Achieved 94% precision, 90% recall, 92% accuracy, and mAP@50/75 of 92%/91%.
• Outperformed baseline models including RetinaNet, DETR, YOLOv10, and DETR v2.
• Demonstrated superior performance in fine-grained detection, especially for rust, bacterial blight, and Fusarium head blight.

Conclusions:

• The proposed model offers significant advantages in theoretical innovation and practical application for intelligent soybean disease detection.
• The endogenous diffusion attention mechanism enhances feature extraction accuracy and robustness.
• Provides critical technological support for precision agriculture and intelligent farming.