Classification of commercial plastics using hyperspectral long wave infrared emission imaging

Mads Nibe Larsen; Anders Løchte Jørgensen; Martin Lahn Henriksen; Victor Petrunin; Jakob Kjelstrup-Hansen; Bjarke Jørgensen; Mogens Hinge

doi:10.1016/j.saa.2024.125501

Classification of commercial plastics using hyperspectral long wave infrared emission imaging

Spectrochim Acta A Mol Biomol Spectrosc. 2024 Nov 27:329:125501. doi: 10.1016/j.saa.2024.125501. Online ahead of print.

Authors

Mads Nibe Larsen¹, Anders Løchte Jørgensen², Martin Lahn Henriksen³, Victor Petrunin², Jakob Kjelstrup-Hansen⁴, Bjarke Jørgensen², Mogens Hinge⁵

Affiliations

¹ NanoSYD, Mads Clausen Institute, University of Southern Denmark, 6400 Sønderborg, Denmark; Newtec Engineering A/S, 5230 Odense, Denmark; Plastic and Polymer Engineering, Department of Biological and Chemical Engineering, Aarhus University, 8200 Aarhus N, Denmark.
² Newtec Engineering A/S, 5230 Odense, Denmark.
³ Plastic and Polymer Engineering, Department of Biological and Chemical Engineering, Aarhus University, 8200 Aarhus N, Denmark.
⁴ NanoSYD, Mads Clausen Institute, University of Southern Denmark, 6400 Sønderborg, Denmark.
⁵ Plastic and Polymer Engineering, Department of Biological and Chemical Engineering, Aarhus University, 8200 Aarhus N, Denmark. Electronic address: [email protected].

PMID: 39626515
DOI: 10.1016/j.saa.2024.125501

Abstract

The long wave infrared regime of the electromagnetic spectrum has been used to identify and classify different types of plastics. This work presents an early-stage version of an industrial hyperspectral camera, which is based on a low-order scanning Fabry-Pérot interferometer and is sensitive to radiation with wavelengths from 8 µm to 15 µm. As a demonstration, the setup is used to record interferograms and distinguish between twelve types of common plastics heated to 60 °C. The plastic identification is based on a nine-component principal component analysis model combined with k-nearest neighbors. The resulting model shows a micro precision of 90 % and a macro specificity of 99 %, being able to classify all tested plastic types except ABS.

Keywords: Fabry-Pérot interferometry; Hyperspectral imaging; Plastics classification; Principal component analysis; Thermography.