Accurate detection of low concentrations of microplastics in soils via short-wave infrared hyperspectral imaging

This study evaluated the effectiveness of coupling machine learning algorithms with short-wave infrared (SWIR) hyperspectral imaging (HSI) in detecting two types of microplastics (MPs) - polyamide (PA) and polyethylene (PE) - with the maximum particle sizes of 50 and 300 μm, respectively, across three batch of concentration ranges (0.01-0.1%, 0.1-1%, and 1-12%) in soils. Using indium gallium arsenide (InGaAs; 800-1600 nm) and mercury cadmium telluride (MCT; 1000-2500 nm) sensors, we applied logistic regression and support vector machines, employing both linear and nonlinear kernels, to analyze spectral features extracted via principal component analysis and partial least squares. The results demonstrated that the overall accuracy for detecting 0.01-12% MPs was 93.8 ± 1.47% using the MCT sensor, which was higher than 68.8 ± 3.76% using the InGaAs sensor. Both sensors showed high accuracy (> 94%) when detecting high levels (1-12%) of MPs in soil, but these accuracies greatly declined as the spiked MP concentrations decreased from 1-12% to 0.1-1% and further to 0.01-0.1%. Moreover, this decline was more pronounced for the InGaAs sensor compared to the MCT sensor and for sub-wavelength spans compared to the full wavelength span under each sensor. The MCT sensor consistently outperformed the InGaAs sensor across all three concentration ranges (0.01-0.1%, 0.1-1%, and 1-12%), potentially due to its extended coverage of 1600-2500 nm and high sensitivity of the detector. Our study highlights the feasibility of the MCT HSI system for rapid and effective detection of MPs in soils non-invasively at concentrations as low as 0.01%. • The MCT sensor outperformed the InGaAs sensor for detecting MPs in soils. • Accuracy for detecting 0.01-12% MPs was 93.8 ± 1.47% using the MCT sensor. • Accuracy for detecting 0.01-12% MPs was 68.8 ± 3.76% using the InGaAs sensor. • Model accuracy declined with the decrease of spiked MP concentrations. • This decline was more obvious for the InGaAs compared to MCT sensor. SynopsisThe overall accuracy for detecting 0.01-12% MPs was 93.8 ± 1.47% using the MCT HSI sensor, greatly higher than 68.8 ± 3.76% achieved with the InGaAs sensor. Both sensors exhibited high accuracies (> 94%) for detecting high MP levels (1-12%) in soils. However, these accuracies declined markedly as the spiked MP concentrations decreased from 1-12% to 0.1-1% and further to 0.01-0.1%.