Taneepanichskul, Nutcha; (2025) Automatic Identification and Classification of Compostable and Biodegradable Plastics using Hyperspectral Imaging. Doctoral thesis (Ph.D), UCL (University College London).

Text Taneepanichskul_10203810_Thesis_id_removed.pdf Access restricted to UCL open access staff until 1 February 2026. Download (8MB)

Abstract

The increasing adoption of compostable and biodegradable plastics, spurred by public and governmental efforts to mitigate plastic waste, underscores the need for efficient sorting technologies to prevent recycling and organic waste streams from contamination. However, despite this urgency, current UK waste management systems often fail to detect and separate compostable plastics, leading to their improper disposal in landfill sites or incinerators. Consequently, this study has focused on developing an effective identification and sorting system for compostable plastics using hyperspectral imaging (HSI) within the short-wave infrared region (SWIR). Employing chemometric techniques like Principal Component Analysis (PCA) and Partial least squares discriminant analysis (PLS-DA) classification models were developed for compostable plastics, including polylactic acid (PLA) and polybutylene adipate terephthalate (PBAT), alongside conventional plastics like polypropylene (PP), polyethylene terephthalate (PET), and low-density polyethylene (LDPE) and compostable material including palm leaf derived packaging and sugarcane derived packaging. Results demonstrated high classification accuracy (>90%) for virgin materials larger than 10 mm x 10 mm. Subsequent enhancement of the PLS-DA classification model through various preprocessing methods—such as combining Savitzky-Golay (SG) (1st derivative, 2nd polynomial, and 15-point window), Standard Normal Variate (SNV), and Mean centering—yielded the highest overall accuracy. This improved model effectively distinguished large microplastic soil-contaminated plastics post in-vessel composting and open windrow, achieving accuracies of 80%. Nonetheless, identification accuracy was shown to be a function of parameters such as darkness, size, colour, thickness, and contamination level. To further enhance model performance in identifying compostable plastics with significant food waste contamination, other machine learning methods were developed such as Support Vector Machine (SVM) methods. These exhibited the high performance in detecting real-world plastic packaging with food contamination. The colour of packaging and contaminant level were found to have the biggest impact on the SVM performance. This research underscores the critical role of advanced sorting technologies in increasing composting and recycling rates, reducing contamination, and fostering a sustainable circular economy.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item