Band Relevance Study of SWIR Hyperspectral Imaging for Material Recycling and Reuse

This paper presents a band relevance analysis for an application of material discrimination with hyperspectral imaging in the context of reusable food and beverage packaging. A high spectral resolution camera in the short wavelength infrared (SWIR) range (1100-1700 nm) is used to extract the relevant bands from the original 100 hyperspectral bands. From the band analysis, based on genetic algorithms, a minimum set of 3 bands can be extracted where high discrimination accuracies can still be achieved. The identified set of relevant bands is valuable input for the future development of quantum-dot based SWIR cameras. These cameras have the potential for lower cost and higher spatial resolution cameras than current InGaAs SWIR cameras, paving the way for wider industrial adoption of SWIR imaging.