Multi-scale closed-loop tuning via spatial frequency collaborative sensitivity for rice leaf disease detection

Area of Science:

• Agricultural Science
• Computer Vision
• Artificial Intelligence

Background:

• Rice is a critical global food source, facing threats from diseases that impact yield and quality.
• Current disease detection relies on subjective methods, leading to yield loss and environmental concerns due to excessive pesticide use.

Purpose of the Study:

• To develop an advanced AI model for early and accurate detection and classification of rice crop diseases.
• To improve the efficiency and sustainability of rice cultivation through precise disease management.

Main Methods:

• Proposed a Multi-scale closed-loop tuning via spatial frequency collaborative sensitivity (MCCA-YOLO) model.
• Integrated a dual-backbone feature extractor, spatial frequency enhancement, and attention mechanisms for robust feature extraction and system self-verification.
• Employed cross-scale weighted fusion and enhanced feature pyramid networks for dynamic adaptation to diverse lesion morphologies.

Main Results:

• Achieved a mean average precision (mAP) of 92.2% on the rice plant diseases v8 dataset, outperforming existing methods.
• Demonstrated high precision (0.915) and recall (0.900) in disease classification.
• Validated superior performance on additional datasets (v9 and RLSD), confirming model robustness.

Conclusions:

• The MCCA-YOLO model offers a significant advancement in early rice disease detection, enhancing agricultural sustainability.
• This AI-driven approach provides a foundation for precision agriculture, optimizing crop management and food security.