Investigation of an Efficient Multi-Class Cotton Leaf Disease Detection Algorithm That Leverages YOLOv11

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Cotton leaf diseases cause significant yield losses.
• Traditional detection methods lack accuracy and are labor-intensive.
• Automated detection is crucial for efficient crop management.

Purpose of the Study:

• To develop an advanced deep learning model for accurate cotton leaf disease detection.
• To address challenges like complex backgrounds, small target detection, and generalization.
• To improve the efficiency and practicality of automated disease monitoring systems.

Main Methods:

• Developed the ACURS-YOLO network based on YOLOv11, integrating U-Net v2 for feature extraction.
• Incorporated CBAM attention mechanism for feature emphasis and SimSPPF for reduced complexity.
• Added C3k2_RCM module for context modeling and ARelu activation function.
• Created a dataset of 3000 cotton leaf disease images and applied data augmentation.

Main Results:

• ACURS-YOLO achieved a mean Average Precision (mAP) of 94.6% (mAP_0.5) and 83.4% (mAP_0.5:0.95).
• The model demonstrated 95.5% accuracy, 89.3% recall, and a 92.3% F1 score with a frame rate of 148 fps.
• Outperformed YOLOv11 and conventional models in detection precision and functionality; ablation tests confirmed component efficacy.

Conclusions:

• The ACURS-YOLO network offers an efficient and accurate solution for automated cotton leaf disease monitoring.
• The integrated modules effectively enhance detection in complex environments.
• This framework advances smart sensor development for agriculture through improved practical applicability.