Complementarity of portable and benchtop near-infrared spectrometers for microplastic predictive detection in whole camel milk powder.

Abstract

This study systematically investigated the complementarity and transferability of benchtop and portable Near Infrared Spectroscopy regarding the detection of common microplastic contamination in camel milk powder (CMP). Both devices achieved superior results in the Support Vector Machine (SVM) discrimination of microplastic contamination types and demonstrated complementarity in the precise identification of pure CMP and its PET contaminated samples. A data level fusion strategy combined with SVM discriminant analysis reduced the classification error of PE, PET, and PS contaminated CMP samples to 0. Benchtop spectrometers were significantly superior to portable devices in the quantitative analysis of microplastic contamination levels, with cross validation coefficients of determination ranging from 0.789 to 0.899 and 0.526-0.717, respectively. Furthermore, the Slope Bias Correction method successfully implemented model transfer; the prediction accuracy of the portable device for PP contaminated samples after correction (R² = 0.796) was improved by 51.10% compared with the original model (R² = 0.526). This study provides a theoretical basis and technical pathway for the rapid onsite screening of microplastic contamination in CMP.