Multispectral Imaging and Hyperspectral Techniques Applied to Dyed Fibers: a Classification Approach

This work investigates the possibility of identification of natural dyes (madder and cochineal) and mordant types (Al, Sn, Cr, Cu, Fe) in dyed wool fibers using spectral imaging-based methods. For this purpose, technical imaging, including UVL, IRR and UVR, and obtained IRFC and UVFC images were used, along with multispectral imaging (350-1100 nm) and hyperspectral imaging (400-950 nm). The grayscales of multispectral and hyperspectral images were extracted to quantify the imaging data. The grayscale and the first derivative of the reflectance spectra obtained from the hyperspectral camera were investigated using multivariate principal component analysis and hierarchical clustering to separate the different groups of dyed fibers. According to results, aluminum and tin mordanted fibers could be distinguished from other groups as per UVL and IRFC images; similarly, the type of dye (madder or cochineal) was distinguishable from UVFC images. Interestingly, PCA analysis of grayscales of multispectral images provided an appropriate separation of all groups of fibers with different mordants and dyes. The 3d PCA plot and the hierarchical clustering of the first derivative of the reflectance spectra also resulted in better separation and classification of the dyed fiber groups. Nevertheless, the best performance in clustering fibers groups can be seen in the PCA analysis of grayscales obtained from hyperspectral images recorded at 430-830 nm. Therefore, hyperspectral imaging can be considered a more appropriate method for categorizing dyed fibers with different dyes and mordants.