On-line monitoring of egg freshness using a portable NIR spectrometer combined with deep learning algorithm

Monitoring and maintaining the freshness of eggs is important to ensuring a supply of eggs that is safe for consumption. Near infrared (NIR) spectrometer has been successfully applied to detect egg freshness. In recent years, a new generation of low-cost, miniaturized NIR sensors has been developed for on-line and in situ food analysis. The purpose of this study is to investigate the performance of a portable NIR spectrometer for on-line evaluation of egg freshness. A deep learning algorithm integrating continuous wavelet transform (CWT) and convolutional neural network (CNN) was proposed to achieve end-to-end prediction of egg freshness and compared with traditional spectral analysis methods based on preprocessing and feature extraction. The results indicated that the proposed CWT-CNN model yielded the optimal performance, with coefficient of determination for prediction (R2P) of 0.9059, root mean square error for prediction (RMSEP) of 4.8153 and residual predictive deviation (RPD) of 3.1201. Furthermore, the identification accuracy of egg freshness grade reached 90.7%, which is comparable to the performance of desktop equipments. This could help food control authorities deploy portable NIR device at different points in the egg supply chain.