Hyperspectral imaging coupled with chemometrics for non-invasive identification and mapping of organic cultural heritage objects

Organic materials are central to many cultural heritage objects, including historical textiles, parchment, and oil paintings. Their layered structure, chemical heterogeneity, and sensitivity to light, heat, and solvents make invasive analysis undesirable, while visual inspection and point techniques alone are often insufficient for spatial interpretation. Hyperspectral imaging provides full field spectra, but its high dimensional data are often noisy, affected by illumination, and contain mixed pixels, which can obscure weak diagnostic features and limit routine use in conservation. This thesis addresses this methodological gap by proposing calibrated, non-invasive workflows that convert hyperspectral image cubes into reliable material maps and quantitative indicators for monitoring change. Three projects couple imaging spectroscopy with chemometric processing to identify organic compounds, map their distributions, and assess degradation or treatment effects. First, the impact of ion beam techniques on protein based substrates (parchment and silk) is evaluated using micro-FTIR mapping with multivariate analysis, supported by confocal micro-Raman spectroscopy to identify degradation markers. Second, varnish cleaning efficiency on oil paintings is quantified using VNIR hyperspectral reflectance imaging and an anchor calibrated spectral angle mapping difference workflow. This converts spectral similarity to clean and varnished references into residue fraction maps and area fraction metrics, enabling spatial comparison of solvent and tissue systems. The method is benchmarked on a laboratory mock-up, tested on a historical painting, and validated by targeted micro-ATR-FTIR measurements on selected mock-up areas. Third, dyes in historical silk textiles, including the Byzantine “Velo di Classe”, are characterized by integrating hyperspectral imaging and micro-Raman spectroscopy to map dye distributions and support identification. Overall, the results demonstrate that hyperspectral imaging combined with chemometrics can deliver reproducible non destructive characterization of organic heritage materials, producing class maps, abundance maps, resolved spectra, and quality metrics that support conservation decisions while minimizing sampling.