Study on the Penetration Depth of Light into Wheat Flour by Hyperspectral Imaging

Abstract

The penetration depth of light into wheat flour is the basis for the effective detection of additives in wheat flour using hyperspectral imaging. To determine the effective penetration depth of light into different gluten flours in hyperspectral image collection, the partial least squares-discriminant analysis (PLS-DA) method was used. Double-layer samples were prepared by placing flour layers with different thicknesses on top of the benzoyl peroxide (BPO) layer. PLS-DA classification model was established by using the diffuse reflectance spectra of each pixel in the double-layer sample image, and the classification accuracy was used to evaluate the results. The results show that the average accuracy of 1 and 1.5 mm models after smoothing pretreatment is above 95%. Therefore, a 1.5 mm sample depth for the detection of mixed samples of flour and additives is recommended. The selected sample depth was used for the detection of mixed samples containing different concentrations of BPO in flour, and the percentage of detected BPO pixels was positively correlated with BPO concentration, which could be used for subsequent quantitative analysis. The results lay a foundation for the effective detection additives in wheat flour by using hyperspectral imaging technology.