Neural Network-based non-destructive quantification of thin coating by terahertz pulsed imaging in the frequency domain

Terahertz pulsed imaging (TPI) is a powerful tool for non-destructive quantification of pharmaceutical tablet coatings. In this paper, we present a Neural Network (NN) based method for extracting the coating thickness from the FFT-amplitude of the measured terahertz waveform. We demonstrate that the NN-based frequency domain method outperforms the standard “peak-finding” time-domain method, in terms of quantifying thinner coating thickness, although a learning set of data is necessary.