On-line monitoring of egg freshness using a portable NIR spectrometer in tandem with machine learning

Despite having an affordable price, several reports of egg mislabeling are published annually, which involves selling stale eggs as fresh. NIR spectroscopy has been successfully used for the prediction of eggs' freshness. In recent years, a new generation of low-cost, portable NIR sensors has been investigated for on-line and in situ food analysis. The main goal of this work was to investigate the performance of one of the smallest and cheapest NIR spectrometer for on-line estimation of egg freshness. Spectral data obtained was processed using different combinations of pre-treatment, and machine learning methods have been assayed to predict the Haugh unit (HU) value (PLS-R and SVM-R) and to classify fresh and stale eggs (PLS-DA and SVM-C). PLS-R and SVM-R regression showed similar performance, but SVM-R model in the spectral region of 1300-1690 nm showed the best results with a relative error of 7.32% and RPD of 2.56. PLS-DA presented better results than SVM-C for the classification of fresh and stale eggs, with an accuracy of 87.0%, with higher sensitivity for identification of stale eggs. The results show that a small portable NIR spectrometer is a cost-effective and reliable device to predict the freshness of hen's eggs with prediction accuracy comparable to benchtop devices. This could help food control agencies implement portable NIR sensors at different egg supply chain stages.

Automated summary

The study evaluated the performance of a small and cost-effective portable NIR spectrometer for estimating the freshness of eggs online. SVM-R model in the spectral region of 1300-1690 nm showed the best results with a relative error of 7.32% and RPD of 2.56. PLS-DA presented better results than SVM-C for the classification of fresh and stale eggs, with an accuracy of 87.0%.