RECYCLING-ORIENTED IDENTIFICATION OF FOSSIL-BASED PLASTICS AND BIOPLASTICS IN PACKAGING WASTE USING HYPERSPECTRAL IMAGING AND HIERARCHICAL CLASSIFICATION

Abstract

The increasing use of bioplastics in packaging leads to their presence within conventional plastic waste streams, where even minor cross-contamination can compromise recycled material quality. Therefore, rapid and efficient material identification strategies are required to support high-purity sorting processes. This study evaluates the applicability of short-wave infrared hyperspectral imaging (SWIR-HSI: 1000–2500 nm) combined with chemometric and machine-learning approaches for the automatic classification of fossil-based plastic and bioplastic flakes from household packaging waste. Five fossil-based polymers commonly used for plastic packaging, i.e., polyethylene terephthalate (PET), high-density polyethylene (HDPE), polypropylene (PP), polystyrene (PS), expanded polystyrene (EPS), and three polylactic acid (PLA)-based bioplastics derived from disposable food packaging items were investigated. Hyperspectral images of the studied samples were acquired in the SWIR range and divided into calibration and validation datasets to develop and evaluate the proposed classification model. After spectral preprocessing, Principal Component Analysis (PCA) was used to explore class separability and guide the development of a hierarchical Partial Least Squares-Discriminant Analysis (Hi-PLS-DA) model. The proposed methodology achieved high predictive performance on the independent validation dataset, with sensitivity and specificity ranging from 0.96 to 1.00. Although conceived as a proof of concept, the proposed approach demonstrates the feasibility of combining SWIR-HSI with hierarchical chemometric classification for recycling-oriented identification of fossil-based plastics and PLA-based bioplastics, supporting the development of more effective sensor-based sorting strategies.