Super-Resolution Imaging Using Very Deep Convolutional Network in Terahertz NDT Field
- Resource Type
- Conference
- Authors
- Wang, Qiang; Zhou, Hongbin; Wang, Yi; Xia, Ruicong; Liu, Qiuhan; Zhao, Boyan; Wang, Yue
- Source
- 2020 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD) Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD), 2020 International Conference on. :438-441 Oct, 2020
- Subject
- Aerospace
Components, Circuits, Devices and Systems
General Topics for Engineers
Robotics and Control Systems
Signal Processing and Analysis
Imaging
Convolution
Testing
Security
Training
Interpolation
Image reconstruction
Terahertz image
Super-Resolution imaging
Convolution network
- Language
In this paper, we propose a Terahertz image Super-Resolution (SR) method based on very deep Convolutional Networks (VDCN) in non-destructive testing (NDT) application. Glass fiber reinforced plastic (GFRP) with prefabrication defects is used as test specimen. Through fast scanning imaging by Terahertz domain spectrometer, we obtained the single original image. Then the HR image can be generated by trained VDCN. Experimental results show that the image restored by our method has higher peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) than traditional Bicubic.