Combining spectroscopy and machine learning for rapid identification of plastic waste: Recent developments and future prospects

Recycling and utilization of plastic waste are receiving more and more attention, and the combination of spectroscopic techniques and machine learning is expected to solve the problem of efficiently identifying and classifying plastic waste at the front end of high-value recycling. Currently, the spectroscopic techniques used for plastic waste classification include near-infrared (NIR) spectroscopy, mid-infrared (MIR) spectroscopy, Raman spectroscopy, laser-induced breakdown spectroscopy (LIBS), X-ray fluorescence (XRF) spectroscopy, terahertz (THz) spectroscopy, etc., and the machine methods combined with them include traditional machine methods and deep learning methods. This paper mainly summarizes the research progress in the application of spectroscopic techniques combined with machine learning in the rapid identification of plastic waste in the past five years, focusing on the innovative research of machine learning methods in plastic identification, the relative advantages and disadvantages of various spectroscopic techniques, and the influencing factors of spectroscopic techniques in plastic identification. In addition, this paper describes the application of spectroscopic instrumentation in the plastic waste recycling industry. In the end, the paper presents an outlook on the future trajectory and potential of this field and proposes recommendations for its advancement in three key dimensions: spectroscopy, machine learning algorithms, and practical engineering applications.