Hyperspectral imaging quantifies blend composition change in workwear textiles

Textile blends are challenging to recycle due to usage of multiple different blend percentages, but also due to composition change caused by fiber degradation over time. This is particularly crucial for workwear, which must meet strict performance and safety requirements. This paper discusses estimating blend composition changes using near infrared hyperspectral imaging. We analyzed 30 used hospital workwear garments with known number of laundering cycles and identical polyester-cotton blend composition at production. A latent variable regression model, based on hyperspectral data, estimated their current composition, which was determined using ISO-standardized chemical analysis. Results showed that near infrared hyperspectral imaging accurately estimated composition changes, with a root mean squared error below 0.5 wt-%, compared to over 1.4 wt-% error when utilizing the number of laundering cycles for estimation. Our approach could be used as a quality control method in sorting, leading to more granular sorted fractions, facilitating increased workwear recycling rates.