Nondestructive Detection of Egg Freshness Fusion Index During Storage Based on Hyperspectral Imaging

Dandan Fu,Qiaohua Wang,Meihu Ma
DOI: https://doi.org/10.1145/3453187.3453379
2020-01-01
Abstract:Hyperspectral transmission images of brown shell eggs were collected by using hyperspectral imager, and the Huff value and yolk index of eggs were measured by electronic balance and vernier caliper respectively. Principal component analysis was used to calculate the weight of egg huff value and yolk index, both of which were 0.5. The two indexes were used to construct the egg freshness fusion index. After comparing the modeling effects of different spectral pretreatments, the full spectrum after the first-order differential pretreatment was selected for subsequent data processing. The characteristic wavelengths which are closely related to the fusion index of egg freshness are selected by using the uninformative variables elimination (UVE), stepwise regression algorithm (STP) and successive projections algorithm (SPA) respectively, and the prediction performance of the model based on the characteristic wavelength is achieved. The results showed that there were 12, 5 and 10 characteristic wavelengths which were closely related to the fusion index of egg freshness selected by UVE, STP and SPA. Among them, the partial least squares (PLS) model established based on 10 characteristic wavelengths screened out by SPA has similar effect to the and multiple linear regression (MLR) model. The correlation coefficients of the PLS model in the training set and the prediction set were 0.860 and 0.856 respectively. The root-mean-square errors were 3.487 and 4.764 respectively, and the RPD of the prediction set was 1.937, which is the optimal model to predict the egg freshness fusion index. The modeling variables were reduced from 438 to 10, the structure of the model was greatly simplified, and the model could predict the egg freshness fusion index non-destructively.
What problem does this paper attempt to address?