DCDW-YOLOv11: An Intelligent Defect-Detection Method for Key Transmission-Line Equipment.

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Defect detection in transmission-line equipment is crucial for grid reliability.
• Complex environments present challenges like background interference and multi-scale features, reducing accuracy.
• Existing methods struggle with precision and robustness in real-world scenarios.

Purpose of the Study:

• To develop an improved defect detection model for transmission-line equipment.
• To enhance accuracy and reliability in complex environmental conditions.
• To provide a robust solution for intelligent transmission line inspection.

Main Methods:

• Proposed DCDW-YOLOv11 model based on YOLOv11.
• Incorporated deformable convolution C2f_DCNv3 for geometric adaptability.
• Integrated Convolutional Block Attention Module (CBAM) for feature highlighting and noise suppression.
• Utilized dynamic head (DyHead) for cross-layer feature fusion and WIoU loss for optimized regression.

Main Results:

• DCDW-YOLOv11 achieved 94.4% accuracy, 92.8% recall, and 96.3% mAP on a transmission-line equipment defect dataset.
• Demonstrated significant improvements over the original YOLOv11 (2.8% accuracy, 7.0% recall, 4.4% mAP).
• Outperformed other mainstream detection models in performance.

Conclusions:

• The DCDW-YOLOv11 model offers high-precision and reliable defect detection capabilities.
• The proposed enhancements effectively address challenges in complex environments.
• The method supports intelligent inspection of transmission lines, improving operational safety and efficiency.