Raman spectroscopy is one of the few non-destructive techniques capable of identifying pigments in art works. Raman spectra contain powerful information that can be used to identify unknown compounds and their chemical structures. However, the analysis of spectral data comes with some difficulties, and therefore the spectral interpretation is not straightforward. Sometimes, there are very little differences in the spectral data concerning to specific identification objectives, for instance, in polymorphic discrimination or in the discrimination of natural and synthetic forms of certain pigments. Moreover, this discrimination is often performed manually so that the process can be repetitive, subjective and particularly time-consuming. The result is an increasing motivation to automate the identification process involved in the classification of pigments in paint. In this paper, we propose a system to automatically classify the spectral data into specific and well-known classes, i.e. reference classes. The proposal is based on a combination of chemometric techniques, which provides a powerful way to achieve spectral separability so that it is possible to discriminate between very similar spectra in an automatic way. In this regard, a decision-making algorithm was specifically developed to select the corresponding reference class with no user input, which was successfully validated using simulated spectra. The implemented methodology was used to classify Raman spectra of pigments commonly present in artist's paints in experimental cases, providing reliable and consistent results. Therefore, the presented system can play a good auxiliary role in the analysts' endpoint classification.

Peer Reviewed