Precise tea leaf disease detection using UAV low-altitude remote sensing and optimized YOLO11 model.

Area of Science:

• Agricultural Science
• Computer Vision
• Remote Sensing

Background:

• Tea leaf diseases significantly impact crop yield and quality.
• Existing intelligent detection methods struggle with complex backgrounds, limited data, and high computational costs.
• There is a need for precise, efficient, and edge-deployable solutions for real-time tea disease monitoring.

Purpose of the Study:

• To develop an improved lightweight detection model, FCHE-YOLO, for accurate and rapid identification of tea leaf diseases.
• To enhance detection accuracy and robustness in complex environments.
• To enable efficient deployment on resource-constrained unmanned aerial vehicle (UAV) edge devices.

Main Methods:

• Proposed FCHE-YOLO model based on YOLO11 with three key optimizations.
• Introduced lightweight backbone module FC_C3K2 to reduce computation and enhance robustness.
• Constructed efficient feature fusion structure HSFPN for multi-scale information integration.
• Designed detection head Efficient Head with group convolution and attention mechanism for improved accuracy.

Main Results:

• FCHE-YOLO improved mean average precision (mAP) from 94.1% to 98.1% compared to YOLO11.
• Inference speed increased by 9.0% (from 43.3 FPS to 47.5 FPS), meeting real-time requirements.
• Computational complexity reduced significantly: FLOPs decreased by 34.3% (6.4 G to 4.2 G), and parameters reduced by 38.9% (2.59 M to 1.46 M).

Conclusions:

• FCHE-YOLO offers superior detection accuracy and reduced missed detections for tea leaf diseases.
• The model's lightweight design and efficiency make it highly suitable for edge deployment on UAVs.
• The developed model is practical for real-time monitoring of tea leaf diseases using UAV-based low-altitude remote sensing.