AUTHOR=Taneepanichskul Nutcha , Hailes Helen C. , Miodownik Mark TITLE=Automatic identification and classification of compostable and biodegradable plastics using hyperspectral imaging JOURNAL=Frontiers in Sustainability VOLUME=4 YEAR=2023 URL=https://www.frontiersin.org/journals/sustainability/articles/10.3389/frsus.2023.1125954 DOI=10.3389/frsus.2023.1125954 ISSN=2673-4524 ABSTRACT=

In the UK waste management systems biodegradable and compostable packaging are not automatically detected and separated. As a result, their fate is generally landfill or incineration, neither of which is an environmentally good outcome. Thus, effective sorting technologies for compostable plastics are needed to help improve composting rates of these materials and reduce the contamination of recycling waste streams. Hyperspectral imaging (HSI) was applied in this study to develop classification models for automatically identifying and classifying compostable plastics with the analysis focused on the spectral region 950–1,730 nm. The experimental design includes a hyperspectral imaging camera, allowing different chemometric techniques to be applied including principal component analysis (PCA) and partial least square discriminant analysis (PLS-DA) to develop a classification model for the compostable materials plastics. Materials used in this experimental analysis included compostable materials (sugarcane-derived and palm leaf derived), compostable plastics (PLA, PBAT) and conventional plastics (PP, PET, and LDPE). Our strategy was to develop a classification model to identify and categorize various fragments over the size range of 50 x 50 mm to 5 x 5 mm. Results indicated that both PCA and PLS-DA achieved classification scores of 100% when the size of material was larger than 10 mm x 10 mm. However, the misclassification rate increased to 20% for sugarcane-derived and 40% for palm leaf-based materials at sizes of 10 x 10 mm or below. In addition, for sizes of 5 x 5 mm, the misclassification rate for LDPE and PBAT increased to 20%, and for sugarcane and palm-leaf based materials to 60 and 80% respectively while the misclassification rate for PLA, PP, and PET was still 0%. The system is capable of accurately sorting compostable plastics (compostable spoons, forks, coffee lids) and differentiating them from identical looking conventional plastic items with high accuracy.