Classification modeling method for hyperspectral stamp‐pad ink data based on one‐dimensional convolutional neural network

Abstract

In the questioned document, the examination of stamp-pad ink is crucial scientific evidence to discern the difference between genuine and forged documents. In this study, a new method for rapid and non-destructive identification of types of stamppad inks by combining hyperspectral imaging (HSI) technology and deep learning was developed. Twenty stamp-pad inks of different brands and models were collected and numbered in turn, and then, each of them was sealed six times repeatedly on the A4 printing paper for the test. After that, the hyperspectral imager was used to collect the hyperspectral images and the reflectance spectral data were obtained after pixel fusion. Principal component analysis (PCA) and non-negative matrix factorization (NMF) were used to deal with the dataset, but visual results were not good. Then, back propagation neural network (BPNN) and one-dimensional convolutional neural network (1D-CNN) were constructed and their merits and drawbacks were compared.The final loss function of the BPNN of training set and validation set was stable at 0.27 and 0.42, and the classification accuracy of the training set and validation set reached 90.02% and 83.99%, respectively. Compared with the BPNN, the 1D-CNN had better stability and efficiency for the classification. The loss function of the training set and validation set was as low as 0.068 and 0.075, and the final classification accuracy reached 98.30% and 97.94%, respectively. Therefore, the combination of hyperspectral imaging technology and 1D-CNN represents a potentially simple, nondestructive, and rapid method for stamp-pad inks detection and classification.

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

“…The advantages of 1D‐CNN for sequence learning should have been significant, such as our previous study in which the classification efficacy for hyperspectral data was far better than that of ordinary machine learning methods. [ 46 ] However, no significant advantage was obtained in this study. It is worth noting that the length of each hyperspectral spectral sequence is only 101, with simple spectral features, while the length of the Raman spectral sequence is 2029, with more complex and variable features.…”

Interpol questioned documents review 2019–2022

“…Stamps have been in use for thousands of years for all forms of business and social use and continue to be important globally in travel and identity documents, as well as regionally with course-of-business documents, Over the period of this review 11 articles [ [503] , [504] , [505] , [506] , [507] , [508] , [509] , [510] , [511] , [512] , [513] ] have been published on this topic, as well as seven conference presentations [ [514] , [515] , [516] , [517] , [518] , [519] , [520] ].…”

“…Wang et al [ 505 ] analyzed 21 stamp-pad ink by hyperspectral imaging (HSI) with chemometric approaches to classify them. From the hyperspectral images, the reflectance spectral data were obtained by pixel fusion.…”

“…In their study, Wang et al [ 511 ] explored a new method for the rapid and non-destructive identification of stamp-pad inks by combining hyperspectral imaging (HSI) technology and deep learning. The authors collected 20 different stamp-pad inks and used them to affix six seals of each on to A4 printing paper.…”

Accurate felt-tip pen brands classification based on a convolutional neural network using data augmentation

“…Identifying the source of inks and colourants can help resolve cases involving documents with a similar appearance. Many studies have analysed and tested the composition [1], colour [2], and brand [3] of ink.…”