Coupling hyperspectral imaging with machine learning algorithms for detecting polyethylene (PE) and polyamide (PA) in...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Spectroscopy

Background:

• Soil is a significant reservoir for microplastic (MP) contamination, especially in agricultural areas.
• Rapid and efficient methods for MP identification in soil are crucial for environmental monitoring.
• Existing methods often require complex digestion and separation procedures.

Purpose of the Study:

• To develop and validate a rapid method for detecting polyethylene (PE) and polyamide (PA) microplastics in soil.
• To evaluate the performance of hyperspectral imaging (HSI) systems coupled with machine learning algorithms for MP identification.
• To determine the minimum detection limits of PE and PA in soil using SWIR HSI.

Main Methods:

• Coupling of Visible Near-Infrared (VNIR), InGaAs, and Mid-Wave Infrared (MWIR) HSI systems with machine learning algorithms.
• Analysis of soil samples spiked with PE and PA microplastics of varying concentrations and sizes.
• Spectral analysis of soil-normalized SWIR spectra to identify unique MP signatures.
• Feature reduction techniques to assess spectral separability.

Main Results:

• SWIR HSI (InGaAs and MCT) demonstrated significant spectral differences between soil and MPs, and among soil-MP mixtures.
• Detection models achieved high accuracies: InGaAs (92-100%) and MCT (97-100%), compared to VNIR (44-87%).
• PE was detected at concentrations as low as 1.6% and PA as low as 5.0% using InGaAs and MCT HSI systems.

Conclusions:

• Hyperspectral imaging (HSI) systems, particularly InGaAs and MCT, are feasible for rapid and accurate detection of PE and PA microplastics in soil.
• The developed method eliminates the need for tedious digestion and separation procedures.
• This approach offers a promising tool for environmental monitoring of microplastic pollution in soils.