Research on Deep Learning-Based Multi-Level Cross-Domain Foreign Object Detection in Power Transmission Lines.

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Deep learning is crucial for power grid safety monitoring, specifically detecting transmission line hazards.
• Existing models face challenges with complex architectures and detecting small or occluded targets, hindering real-time and edge-device applications.

Purpose of the Study:

• To develop an efficient and accurate deep learning model for detecting potential safety hazards in transmission lines.
• To address limitations of existing models in detecting small or occluded targets and for edge-device deployment.

Main Methods:

• Integration of the YOLOv11 model with the ConvNeXt network to create the ConvNeXt-You Only Look Once (CO-YOLO) model.
• Utilized Bayesian optimization for hyperparameter tuning to accelerate model convergence.

Main Results:

• CO-YOLO achieved a mean average precision (mAP) of 98.4% at IoU threshold 0.5 and 66.1% at IoU threshold 0.5:0.95.
• The model demonstrated a frames per second (FPS) of 303, outperforming YOLOv11 and ETLSH-YOLO in both accuracy and efficiency.
• CO-YOLO showed a 1.9% improvement in mAP@0.5 and a 2.2% improvement in mAP@0.5:0.95 compared to the original YOLOv11 model.

Conclusions:

• The proposed CO-YOLO model significantly enhances the accuracy and efficiency of transmission line safety hazard detection.
• CO-YOLO is a promising solution for real-time monitoring and edge-device deployment in power grid safety applications.