A Novel NIR-Based Strategy for Rapid Freshness Assessment of Preserved Eggs

The flavor of many traditional Chinese foods, such as preserved eggs, improves with increasing storage time. However, prolonged storage can also have hidden dangers for human health. Therefore, rapid, high-efficiency, contaminant-free, and non-destructive assessment of preserved egg freshness is essential to ensure food safety. The total volatile base nitrogen can be used as an indicator of freshness of preserved eggs. In this study, the near infrared (NIR) spectroscopy was combined with chemometrics for quick determination of preserved egg freshness. The Savitzky-Golay smoothing algorithm was applied for pretreatment, while principal component analysis (PCA) was adopted for outlier removal from samples. NIR models were built by applying the partial least squares regression and support vector regression (SVR) techniques. The result showed that, compared to the other pretreatment methods, multi-scatter correction and standard normal variate (SNV) delivered the best spectral processing effects, and PCA effectively identified and eliminated outliers by confidence ellipse from preserved egg samples. The calibration set spectral matrix processed by SNV, PCA, and interval random frog is combined to construct the SVR model. Prediction set spectrum is substituted into the SVR model, and the predicted TVB-N is compared with the true value, which produced the coefficient of determination of SVR model is 0.91, root mean square error of 1.30, and residual prediction deviation of 3.47, respectively. The results achieved herein showed that by combining the NIR spectroscopy with chemometrics, the freshness of preserved eggs can be effectively assessed; therefore, it can be a useful complementary method for food safety testing.

Automated summary

This study combines near infrared (NIR) spectroscopy with chemometrics to establish a model for quick determination of preserved egg freshness. The results show that multi-scatter correction and standard normal variate (SNV) processing methods provide the best spectral processing effects, and principal component analysis (PCA) effectively removes outliers. By combining the processed spectral matrix of the calibration set, a support vector regression (SVR) model is constructed to accurately predict the freshness of preserved eggs.