Advancing hyperspectral imaging for cultural heritage: new acquisition systems and multivariate image analysis

Cultural heritage serves as a bridge between historical eras and the present, providing invaluable insights into the evolution of human civilisation and artistic expression. However, the complexity and diversity of materials, coupled with their irreversible degradation over time, constantly challenge scientific research to characterize materials used in cultural heritage, addressing conservation and restoration issues. This dissertation explores advancements in hyperspectral imaging (HSI) systems and chemometric data processing methods for analysing cultural heritage materials non-invasively. The research investigates and enhances the application on several HSI systems: 1.The IRIS system by XGLab: enables the co-registered acquisition of XRF (0-48 keV), VNIR (380–1100 nm) and SWIR (1100–2500 nm) reflectance spectroscopy, and excels in the in-depth analysis of multi-layered paintings. 2.A custom-made ER-FT-IR mapping system: combines a portable FT-IR spectrometer (5500-650 cm-1) with a high-precision motorised stage to explore the capabilities of macro-reflectance mapping in the NIR-MIR region for the investigation of multi-layered paintings. 3.A steady-state HSI system integrates reflectance (400-1000nm) and fluorescence (540-1000nm) imaging for analysis of dyed wool samples, focusing on dye identification and degradation detection. Given the massive and complex data generated by the HSI, parallel research is being conducted into chemometrics based multivariate image analysis. Several explorations have been carried out, adapted to different data generated by different systems, including data fusion, PCA, wavelet denoising, clustering methods, and MCR-ALS for efficient data reduction, information extraction, interpretation, and visualisation. These techniques are adept at analysing complex painting compositions, stratigraphic information, and recognise the degradation patterns and stages of dye degradation in wool samples. Overall, this dissertation aims to deepen the understanding of the sophisticated relationship between hyperspectral imaging techniques and data processing explorations, stressing their essential role in the advanced analysis of cultural heritage materials.