Using hyperspectral imaging and machine learning to identify food-contaminated compostable and recyclable plastics.

Area of Science:

• Materials Science
• Environmental Science
• Computer Science

Background:

• Compostable plastics offer an alternative to conventional plastics for food packaging.
• Effective waste management requires distinguishing compostable plastics from other waste streams, especially when contaminated with food waste.
• Current near-infrared technology struggles to identify plastics contaminated with food waste.

Purpose of the Study:

• To investigate the application of hyperspectral imaging for detecting and sorting compostable plastics contaminated with food waste.
• To develop and evaluate machine learning algorithms for accurate classification of contaminated compostable plastics.
• To assess the impact of plastic features on detection model performance.

Main Methods:

• Hyperspectral imaging was employed to capture spectral data from plastic samples.
• Various machine learning algorithms were combined with hyperspectral data for classification.
• The influence of plastic characteristics like darkness, size, and contamination level on model accuracy was analyzed.

Main Results:

• Hyperspectral imaging combined with machine learning achieved up to 99% accuracy in identifying compostable plastics with food waste contamination.
• Plastic darkness was identified as the most significant factor affecting model performance.
• The developed model demonstrated improved accuracy in detecting plastics with higher contamination levels compared to previous studies.

Conclusions:

• Hyperspectral imaging is a promising technology for enhancing the detection and sorting of contaminated compostable plastics in waste streams.
• Implementation in waste management systems can significantly increase composting and recycling rates.
• This approach supports the circular economy by improving the processing and quality of recycled materials.