Discernment of textile fibers by polarization-sensitive Digital Holographic microscope and machine learning

Garment quality and preciousness depend on the type of textile fiber used in the manufacturing. The softer and rarer the animal fiber, the more expensive the textile garment. The cheapest clothes are made by mixing precious fibers such as cashmere with common ones such as sheep wool. To stop clothing counterfeit and quality forgery, checking the type of animal fibers used in textile industries is pivotal. More in general, law regulations require that the declared composition of a tissue meet some standards of quality that have to be assayed carefully by expert operators. Microscopy techniques such as Scanning Electron Microscopy (SEM) and Light Microscopy (LM) are commonly used to discriminate between textile animal fibers. However, analysis by SEM and LM depends on skilled experts called to judge, one-by-one, each fiber. This process is slow, cumbersome, and may be inaccurate, especially if the textile fibers share similar morphologies. Furthermore, the chemical treatments required by some textile processes can heavily modify the morphology of the fibers making more difficult to get correct results. In this work, the textile animal fibers are characterized by a polarization-sensitive, stain-free, Digital Holographic Microscopy (DHM) technique. In particular, we show how cashmere and wool fibers differ according to their anisotropy properties, e.g., birefringence. The optical characterization of textile fibers through the Jones matrix formalism allowed us extracting polarization-dependent DH features capable of accurately classifying three types of animal microfibers using a machine learning approach. Such promising results smooth the path towards an automatic, rapid, and objective identification process for textile industry and standardization purposes.