An Improved YOLOv8 Network for Detecting Electric Pylons Based on Optical Satellite Image

Area of Science:

• Computer Vision
• Artificial Intelligence
• Electrical Engineering

Background:

• Accurate detection of electric pylons is vital for monitoring power transmission lines and infrastructure.
• Existing optical satellite image-based models face challenges in detecting electric pylons due to their unique characteristics and complex environments.

Purpose of the Study:

• To propose an innovative deep learning model, EP-YOLOv8, for enhanced electric pylon detection in optical satellite imagery.
• To improve the accuracy and efficiency of electric pylon identification for better infrastructure management.

Main Methods:

• Development of the EP-YOLOv8 network, an optimized version of YOLOv8n.
• Integration of novel modules: DSLSK-SPPF for enhanced feature capture and EMS-Head for detailed, lightweight detection.
• Evaluation using mAP@0.5 metric.

Main Results:

• The EP-YOLOv8 network achieved a high average mAP@0.5 of 95.5% for electric pylon detection.
• The DSLSK-SPPF and EMS-Head modules effectively captured surrounding features and fine details, respectively.
• Demonstrated superior performance over existing models in detecting electric pylons.

Conclusions:

• The EP-YOLOv8 network offers a significant advancement in electric pylon detection accuracy.
• This model enhances the monitoring of power infrastructure's operational status and layout.
• Provides crucial insights for effective infrastructure management and maintenance.