Segmenting the Semi-Conductive Shielding Layer of Cable Slice Images Using the Convolutional Neural Network

Area of Science:

• Electrical Engineering
• Materials Science
• Computer Vision

Background:

• The semiconductive shielding layer is crucial for aerial insulated cable performance.
• Accurate measurement of structural parameters is essential for cable quality.
• Image processing is vital for these measurements, but cutting marks cause significant interference.

Purpose of the Study:

• To propose a novel convolutional neural network (CNN) based method for accurate image segmentation of semiconductive shielding layers.
• To address the challenge of image noise, specifically cutting marks, that hinder structural parameter measurements.
• To enhance the reliability and precision of cable quality assessment through improved image analysis.

Main Methods:

• A deep fully convolutional network (FCN) framework with skip connections was developed.
• Inception structures and residual connections were integrated for multi-scale feature fusion.
• An improved weighted loss function and a refined pixel classification algorithm were employed.

Main Results:

• The proposed CNN method demonstrated superior performance in image segmentation compared to existing algorithms.
• The technique effectively mitigated the impact of cutting marks on image analysis.
• Accurate segmentation led to more reliable structural parameter measurements.

Conclusions:

• The novel CNN-based image segmentation method offers a significant advancement for analyzing semiconductive shielding layers in aerial insulated cables.
• This approach enhances the precision of structural parameter measurements, contributing to improved cable quality and transmission efficiency.
• The developed algorithm provides a robust solution for overcoming image disturbances in critical cable component analysis.