HCDFI-YOLOv8: A Transmission Line Ice Cover Detection Model Based on Improved YOLOv8 in Complex Environmental...

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Unmanned aerial vehicles (UAVs) face challenges in transmission line ice detection due to variable angles and complex backgrounds.
• Existing methods suffer from poor ice-covering recognition accuracy and target identification difficulties.

Purpose of the Study:

• To develop an improved icing detection model for transmission lines using UAVs.
• Enhance the accuracy and robustness of ice cover detection in challenging environments.

Main Methods:

• Proposed an improved You Only Look Once version 8 (HCDFI-YOLOv8) model.
• Introduced a cross-dense hybrid (CDH) parallel heterogeneous convolutional module for improved accuracy and computational efficiency.
• Implemented deep and shallow feature weighted fusion with an improved CSPDarknet53 to 2-Stage FPN_Dynamic Feature Fusion (C2f_DFF) module to reduce feature loss.
• Optimized the detection head with a feature adaptive spatial feature fusion (FASFF) module for multi-scale feature extraction.
• Developed a new inner-complete intersection over union (Inner_CIoU) loss function to address limitations of the original CIOU loss.

Main Results:

• The HCDFI-YOLOv8 model achieved a 2.7% improvement in mAP@0.5 and a 2.5% improvement in mAP@0.5:0.95 compared to standard YOLOv8.
• Demonstrated superior overall detection accuracy among twelve evaluated models for icing detection.
• Verified the model's effectiveness in complex transmission line environments.

Conclusions:

• The proposed HCDFI-YOLOv8 model significantly enhances transmission line ice cover detection accuracy.
• The model provides effective technical support for UAV-based ice detection in challenging conditions.
• The innovations in convolutional modules, feature fusion, and loss function contribute to superior performance.