Research on identification of ink marks based on machine learning and laser-induced breakdown spectroscopy

Abstract

In recent years, new technologies are emerging in the field of judicial expertise, followed by more arduous challenges. In this study, ink marks are used as an example. Meanwhile, machine learning and laser-induced breakdown spectroscopy (LIBS) are used to analyze the ink marks. This is a new idea in the field of handwriting identification. First, the spectrum is obtained by LIBS. The characteristic spectral lines of C, N, O, Si, Mg, Al, and Ca are observed in the spectrum. Second, a detailed spectrum of the ink mark is provided in this article; in addition, different kinds of inks are used for analogy observation to analyze the influence of different components on ink marks. Finally, the K-nearest neighbor algorithm based on the principal component analysis is used to build the ink recognition model and then analyze the differences in different inks and build a database. The identification results become more intuitive and accurate combining machine learning based on big data, which provide reliable evidence for judicial expertise.