Visualization research of egg freshness based on hyperspectral imaging and binary competitive adaptive reweighted sampling

Abstract

Freshness is an important indicator for evaluating egg quality and is crucial for the food processing industry and consumers. The aim of this study is to non-destructively detect and visualize the freshness of eggs during storage by using hyperspectral imaging (HSI) and multivariate analysis. The hyperspectral images of egg samples with different storage time were collected in the spectral range of 401–1002 nm. A binary competitive adaptive reweighted sampling (BCARS) algorithm considering the synergetic effect among variables was proposed to select feature wavelengths from the whole spectral range and compared with competitive adaptive reweighted sampling (CARS). A slime mould algorithm optimized support vector regression (SMA-SVR) model was proposed to develop calibration models for HU (egg freshness indicator). Statistical analysis results indicated that the proposed BCARS had better feature extraction performance than CARS and the SMA-SVR model outperformed the compared models, in which the BCARS-SMA-SVR model yielded the best performance with a determination coefficient (R2) of 0.946 for calibration set and 0.914 for prediction set. Finally, by transferring the quantitative model to each pixel of hyperspectral image, the visualizing distribution map of HU was generated, providing an intuitive evaluation for egg freshness, which facilitates to the management of storage and marketing. The results provided the possibility of implementing a multispectral imaging for online monitoring of egg quality.