A. Sarker, S. Singh, M. Salas, R. Rao, L. Pal
North Carolina State University,
United States
Keywords: hyperspectral imaging, colored plastic waste, machine learning, real time characterization
Summary:
Plastics containing colorants, fillers, and additives pose significant challenges for optical sorting systems due to their altered reflectance and absorption characteristics, which make them difficult to detect using conventional near-infrared (NIR) sensors. This study focuses on the spectral and chemometric characterization of these hard-to-detect plastics (e.g., colored or black plastics) using visible–near-infrared (VNIR, 400–1000 nm) and shortwave infrared (SWIR, 1000–2500 nm) hyperspectral imaging (HSI). A spectral library of representative polymer types, including high density polyethylene (HDPE), polypropylene (PP), polystyrene (PS), and polyethylene terephthalate (PET), incorporating pigments, colorants, dyes, fillers, stabilizers, and other additives, was developed. Spectral preprocessing, principal component analysis (PCA), and machine learning (ML)–based feature extraction were applied to identify diagnostic absorption features associated with polymer backbones and additive interactions. The resulting dataset enables AI–based classification models for real-time waste material characterization, particularly for colored plastics, establishing a framework for integrating hyperspectral imaging and data analytics to enhance the recycling of complex, additive-rich plastic waste streams.