Nondestructive optical and spectroscopic techniques combined with machine learning for identifying solid waste: A review

Abstract

Characterizing solid waste is crucial for effective waste management strategies. Traditional waste analysis methods are laborious, time-consuming, and destructive, prompting the emergence of nondestructive techniques. The adoption of efficient machine learning and deep learning methods can enhance data processing and recognition accuracy. However, previous reviews have focused on individual analytical techniques, lacking comprehensive comparisons and systematic summaries. This review aims to compare and discuss rapid and nondestructive optical and spectroscopic techniques, including computer vision, vibrational spectroscopy, hyperspectral imaging, and thermal imaging, combined with the latest developed algorithms to improve solid waste identification and characterization. These techniques have been applied in waste classification and sorting, detecting macro and microplastics, recognizing harmful contamination, and predicting chemical properties. While demonstrating high performance, limitations and challenges remain. Future research is necessary on dual-sensor platforms and deeper exploration of solid waste properties. This review will guide the advancement of nondestructive techniques for rapid solid waste recognition.