Rapid detection of mussels contaminated by heavy metals using near-infrared reflectance spectroscopy and a constrained difference extreme learning machine

Abstract

The consumption of mussels contaminated with heavy metals can cause toxicity in humans. To realize quick, accurate, and non-destructive detection of heavy metals in mussels, a new method based on near-infrared reflection spectroscopy was developed in this study. Spectral data from 900 nm to 1700 nm of non-contaminated mussels and mussels contaminated with Cd, Zn, Pb, and Cu were collected using a near-infrared spectrometer. After pre-processing spectral data with multiplicative scatter correction, wavelength selection algorithms based on consistency measures of neighborhood rough sets were used to extract wavelengths for distinguishing non-contaminated and contaminated mussels. A constrained difference extreme learning machine was established as a classification model to detect contaminated mussels. In the proposed model, the weight and bias of the hidden layers are calculated by the difference vectors of samples between classes instead of being randomly selected. The results indicate that the proposed model performs significantly well in differentiating between non-contaminated and contaminated mussels. The average classification accuracy of 50 randomly generated test datasets reaches 97.53%, 95.67%, 99.00%, and 98.80% for the detection of Zn, Pb, Cd, and Cu contamination, respectively. This study demonstrates that near-infrared spectroscopy coupled with a constrained difference extreme learning can be used to rapidly and accurately detect mussels contaminated with heavy metals. This is of great significance for the evaluation of the quality and safety of mussels.