Innovative segmentation technique for aerial power lines via amplitude stretching transform

Area of Science:

• Electrical Engineering
• Computer Vision
• Image Processing

Background:

• Accurate power line segmentation is crucial for fault detection and power system stability.
• Unmanned Aerial Vehicle (UAV) imagery presents challenges due to small target size, weak signal, and environmental noise, leading to low accuracy and efficiency in existing methods.

Purpose of the Study:

• To design an advanced algorithm for accurate and efficient segmentation of aerial power lines from UAV images.
• To overcome limitations of current methods in detecting power lines amidst environmental interference and noise.

Main Methods:

• A pure amplitude stretching kernel function was designed to create a Fourier amplitude vector field.
• This vector field was used to transform the amplitude field of aerial power line images.
• Relative Total Variation (RTV) processing was applied for final power line detection.

Main Results:

• The proposed algorithm achieved high average evaluation indicator scores: P_PA=0.96, M_MPA=0.96, and M_MIoU=0.95.
• Detection time lag averaged less than 0.2 seconds, demonstrating superior real-time performance.
• The method significantly outperformed established algorithms like Region Convolutional Neural Networks (R-CNN) and Phase Stretch Transform (PST).

Conclusions:

• The developed algorithm offers significantly improved accuracy and real-time efficiency for aerial power line segmentation.
• This technique provides a robust solution for enhancing the safety and stability of power systems through better image analysis.