Rapid detection of microplastics in chicken feed based on near infrared spectroscopy and machine learning algorithm

The main objective of this study was to evaluate the potential of near infrared (NIR) spectroscopy and machine learning in detecting microplastics (MPs) in chicken feed. The application of machine learning techniques in building optimal classification models for MPs-contaminated chicken feeds was explored. 80 chicken feed samples with non-contaminated and 240 MPs-contaminated chicken feed samples including polypropylene (PP), polyvinyl chloride (PVC), and polyethylene terephthalate (PET) were prepared, and the NIR diffuse reflectance spectra of all the samples were collected. NIR spectral properties of chicken feeds, three MPs of PP, PVC and PET, MPs-contaminated chicken feeds were firstly investigated, and principal component analysis was carried out to reveal the effect of MPs on spectra of chicken feed. Moreover, the raw spectral data were pre-processed by multiplicative scattering correction (MSC) and standard normal variate (SNV), and the characteristic variables were selected using the competitive adaptive re-weighted sampling (CARS) algorithm and the successive projections algorithm (SPA), respectively. On this basis, four machine learning methods, namely partial least squares discriminant analysis (PLSDA), back propagation neural network (BPNN), support vector machine (SVM) and random forest (RF), were used to establish discriminant models for MPs-contaminated chicken feed, respectively. The overall results indicated that SPA was a powerful tool to select the characteristic wavelength. SPA-SVM model was proved to be optimal in all constructed models, with a classification accuracy of 96.26% for unknow samples in test set. The results show that it is not only feasible to combine NIR spectroscopy with machine learning for rapid detection of microplastics in chicken feed, but also achieves excellent analysis results.