Black plastic identification by hyperspectral imaging in mid-wave infrared

Area of Science:

• Materials Science
• Analytical Chemistry
• Environmental Science

Background:

• Near-Infrared (NIR) sensor-based sorting is standard for post-consumer plastics.
• NIR struggles with black plastics due to carbon black pigment absorption.
• Mid-Wave Infrared (MWIR) offers an alternative wavelength range for analyzing black polymers.

Purpose of the Study:

• To compare the efficacy of MWIR versus NIR hyperspectral imaging for plastic waste classification.
• To evaluate performance on both black and colored plastic waste.
• To assess various chemometric methods, including convolutional neural networks (CNNs).

Main Methods:

• Collected spectral data from five common polymers: HDPE, LDPE, PET, PP, and PS.
• Utilized MWIR and NIR hyperspectral imaging techniques.
• Applied chemometric methods, notably CNNs, for spectral classification.

Main Results:

• MWIR demonstrated superior performance for black plastic classification (83.4% balanced accuracy with CNN) compared to NIR (47.5%).
• NIR showed a slight advantage for colored plastics, with 7 percentage points higher balanced accuracy than MWIR.
• CNN outperformed other chemometric methods across all sensors and sample types.

Conclusions:

• MWIR hyperspectral imaging is a highly effective alternative to NIR for sorting plastic waste, particularly in high-black-plastic scenarios.
• This technology can significantly improve the recycling rates of black plastic packaging.
• The findings support the integration of MWIR imaging into waste management infrastructure.