MULTISPECTRAL OPTICAL SENSORS FOR INTEGRATED WASTE MANAGEMENT

This research presents two innovative approaches for efficient and cost-effective material classification using SWIR (Short-Wave Infrared) spectroscopy, targeting applications in smart waste bins and small-scale industrial systems. The first approach employs SWIR discrete spectroscopy in reflectance mode with integrated LED light sources and a Ge photodetector, achieving up to 98% classification accuracy for plastics and non-plastics. The second approach introduces a voltage-tunable Ge-on-Si photodetector combined with supervised and semi-supervised machine learning algorithms. Operating with a halogen light source and advanced signal enhancement tools, this system enables accurate real-time classification across visible and SWIR spectra. It demonstrates 95% accuracy using few voltage steps, ideal for fast, on-site waste sorting. Both methods leverage compact, energy-efficient optical components and advanced data processing to offer scalable solutions for material identification. The findings underscore the potential of these systems to improve recycling efficiency and support sustainable waste management through intelligent, high-performance sensing technologies.