Scholarlink

This research focuses on the development of an automatic system to classify disposable waste, based on the combination of a dual-band photodetector array operating in visible and short-wave-infrared wavelengths with machine learning algorithms. The identification is performed through spatially resolved spectral reflectivity of real waste samples for a total of seven classes: four plastic polymers (PET, HDPE, LDPE, PP), paper, aluminum and glass. Several classification algorithms were tested, with the best results obtained using Bagged Trees model combined with preprocessing and features selection techniques. Accuracies of approximately 95 and 80% were achieved for three (PET, HDPE, PP) and seven classes, respectively.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Realization of an Automatic Material Recognition System for Waste Recycling

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Claudio Fratini,
  • Rahmi Elagib,
  • Andrea De Iacovo,
  • Andrea Ballabio,
  • Jacopo Frigerio,
  • Giovanni Isella,
  • Andrea Ria,
  • Paolo Bruschi,
  • Lorenzo Colace

摘要

This research focuses on the development of an automatic system to classify disposable waste, based on the combination of a dual-band photodetector array operating in visible and short-wave-infrared wavelengths with machine learning algorithms. The identification is performed through spatially resolved spectral reflectivity of real waste samples for a total of seven classes: four plastic polymers (PET, HDPE, LDPE, PP), paper, aluminum and glass. Several classification algorithms were tested, with the best results obtained using Bagged Trees model combined with preprocessing and features selection techniques. Accuracies of approximately 95 and 80% were achieved for three (PET, HDPE, PP) and seven classes, respectively.