Terahertz image super-resolution reconstruction method for polyethylene pipe heat-fused joint defects based on an...

Area of Science:

• Materials Science
• Non-destructive Testing
• Artificial Intelligence

Background:

• Terahertz imaging offers potential for non-destructive defect detection in polyethylene pipelines.
• Current deep learning super-resolution methods struggle with artifacts in low-resolution terahertz images, hindering defect analysis.
• Efficient and accurate defect identification is crucial for pipeline integrity.

Purpose of the Study:

• To develop an enhanced super-resolution generative adversarial network (SRGAN) for terahertz images.
• To improve the reconstruction quality and artifact reduction in low-resolution terahertz defect detection.
• To enhance the detection of defects like holes and cracks in polyethylene pipeline fusion joints.

Main Methods:

• Proposed an enhanced SRGAN incorporating an efficient multi-scale attention module in the generator.
• Implemented a vision transformer-based discriminator to improve global image consistency discernment.
• Utilized a composite loss function to guide edge information restoration.

Main Results:

• The enhanced SRGAN achieved superior performance in Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) compared to five mainstream methods.
• The model demonstrated improved handling of image texture and high-frequency details.
• Effectively enhanced the reconstruction quality for terahertz images of hole and crack defects.

Conclusions:

• The proposed method significantly improves terahertz image super-resolution reconstruction quality for defect detection.
• The enhanced SRGAN shows strong potential for reliable defect identification in polyethylene pipeline fusion joints.
• This approach offers a more efficient and accurate solution for non-destructive testing in industrial applications.