Optimization technique for transmissive THz-TDS detection in non-destructive testing conventional environments based on filtering and a peak attention network

Terahertz time-domain spectroscopy (THz-TDS) is an advanced non-destructive technique that performs quality inspection of non-polar and non-metallic materials. However, THz-TDS is susceptible to high water vapor content and system noise in non-destructive testing conventional environments. In this study, we developed a THz-TDS detection optimization technique based on a filter peak attention network (FPAN) that integrates filtering, pooling, and peak attention mechanisms to achieve high signal-to-noise ratio data from a single measurement in non-destructive testing conventional environments. The proposed FPAN outperforms the existing methods in all the indicators, effectively removing the water vapor interference and system noise and improving the detection speed.