Classification of black plastic types by hyperspectral imaging based on long-wave infrared emission spectroscopy

Identification of black plastics poses a significant challenge in recycling due to the absorptive nature of carbon black additives. This work introduces a method where hyperspectral imaging in the long-wave infrared regime is used to distinguish between 12 samples of commercially available black plastics encompassing nine distinct polymer types. The spectral scanner comprises a scanning Fabry-Pérot interferometer and a thermal camera based on an uncooled microbolometer detector sensitive to wavelengths from 8 μm to 15 μm. A principal component model is combined with k-nearest neighbors to differentiate between plastic samples in hyperspectral images. The model successfully classifies five (PET, POM, PMMA, PA6, and PA66) out of nine black polymers, and the overall accuracy of the model is %. • Plastic type classification of black plastics was obtained by emission spectroscopy • The plastic analysis is obtained via low-order Fabry-Pérot interferometry • Data cubes from hyperspectral imaging with 155 spectral channels are recorded • A narrow band IR filter is applied as external spectral reference • Principal component analysis and K-nearest neighbours is applied for classification