Search

[Objectives] To discuss the possibility of the non-destructive prediction of the total viable count of bacteria in the polluted eggs, this paper explored a non-destructive method that using near infrared hyperspectral imaging to predict and visualize the total viable count of bacteria in eggs. [Methods] Egg samples were inoculated with the mixture solution of Escherichia coli and Pseudomonas aeruginosa, then the original hyperspectral information and the total viable count of bacteria of eggs in different pollution levels during storage were collected. After selecting the best pretreatment method in this experiment, the predicting model by partial least squares (PLS) and support vector machine (SVM) model were developed based on full wavelengths and the feature wavelengths via the successive projections algorithm (SPA). Finally, the relative best prediction model of the viable count of bacteria was selected and the visualization study of the bacteria was realized in eggs. [Results] The second derivative method was the relative best pretreatment method, the correlation index of cross validation (RCV) was 0.88, and the root mean square error of cross validation was 0.82 lg (CFU·g-1). The relative best prediction model was the SVM model which was based on the feature wavelengths, the correlation index of calibration (RC) and prediction (RP) was 0.88 and 0.84, and the root mean square error of calibration and prediction was 0.86 and 0.97 lg (CFU·g-1), respectively. The visualized images of the polluted levels in eggs were built based on the difference of the pollution levels and spectral characteristics. [Conclusions] The near infrared hyperspectral imaging technology, and multivariate statistical and image processing methods can realize the prediction and visualization of the total viable count of bacteria in eggs, and this technology can also provide the foundation for on-line detection of the egg security. [ABSTRACT FROM AUTHOR]