Classification of commercial plastics using hyperspectral long wave infrared emission imaging

Area of Science:

• Spectroscopy
• Materials Science
• Optics

Background:

• Long wave infrared (LWIR) spectroscopy is effective for plastic identification.
• Hyperspectral imaging offers detailed spectral information for material classification.

Purpose of the Study:

• To present an early-stage industrial hyperspectral camera prototype.
• To demonstrate the camera's capability in classifying common plastics using LWIR radiation.

Main Methods:

• Utilized a scanning Fabry-Pérot interferometer sensitive to 8-15 µm wavelengths.
• Recorded interferograms of twelve common plastic types heated to 60°C.
• Applied principal component analysis (9 components) and k-nearest neighbors for classification.

Main Results:

• Successfully distinguished between twelve common plastic types.
• Achieved 90% micro-precision and 99% macro-specificity.
• The classification model was unable to identify ABS plastics.

Conclusions:

• The developed hyperspectral camera shows promise for industrial plastic identification.
• LWIR spectroscopy combined with PCA and k-NN is a viable method for plastic classification.
• Further development is needed to improve the identification of all plastic types, including ABS.