Global textile production is mainly based on polyester and cotton fibers. A majority of textiles at the end of their lifecycle are currently landfilled or incinerated, but will be increasingly recycled in the future. Here, we discuss how the polyester content in blended textiles can be estimated based on hyperspectral near infrared imaging with the aim of developing machine vision for textile characterization and recycling. Differences in the textile samples were first visualized based on a principal component model and the polyester contents of individual image pixels were then predicted using image regression. The results showed average prediction errors of 2.2-4.5% within a range of 0-100% polyester and enabled visualizing the spatial changes in the polyester contents of the textiles. We foresee that digitalized tools similar to what we report here will be increasingly important in the future as more emphasis is placed on coordinated collection, sorting and reuse of waste textiles.
