The development of a joint neural network to the Raman spectroscopy identification of stamp‐pad ink brands and aging

Abstract The identification of seal authenticity is an important part of document inspection. Confocal laser Raman spectroscopy combined with convolutional neural (CNN) and recurrent neural (RNN) networks was used to distinguish red stamp‐pad ink of different brands and aging. A total of 536 spectral samples from 16 brands were collected in this study and 53,600 amplification data samples were obtained by adding noise. The joint neural network has significant classification performance compared to partial least squares (PLS) discriminant and common CNN. In the three kinds of stamp‐pad inks, photosensitive, atomic, and common, the recognition rate for different brands and aging both reached 100%.