RESUMEN
In this work samples of historical pigments of green hue were brushed on a canvas and studied by Visible Reflectance, X-Ray Fluorescence and Fourier Transform Infrared Spectroscopy. One hundred samples were investigated, all with green hue, these prepared from pigments themselves green, such as chromium oxide (Cr2O3) or from a mixture of pigments that result in green, for example, chrome yellow (PbCrO4) and Prussian blue (Fe4[Fe(CN)6]3). Because every sample investigated through the spectroscopic techniques were of green hue, the characterization of the pigments present in the mixtures through the visual inspection of spectra has become a complex task in some cases, also, due the large number of recorded spectra. In this work, classification models were developed using the multivariate statistical method Partial Least Squares Discriminant Analysis (PLS-DA) to automate the characterization of the pigments present in the mixtures. The models were developed to classify chromium oxide (Cr2O3), chrome yellow (PbCrO4), cerulean blue (CoO.nSnO2) and yellow ochre (Fe2O3·H2O + clay + silica). The models were developed from the fusion of data from the three spectroscopic techniques. However, before data fusion, pre-treatments of the spectral data were tested for their influence on the PLS-DA models. The models developed with data from the three techniques made it possible to classify the pigments of interest in the samples with up to 100% effectiveness. The results also indicate that fusion of the data from the three techniques allows to obtain fingerprints of the pigments of interest, which is not always possible using data from only one or two of the techniques applied in this work.