Growth Identification of Penicillium by Visible/Near-Infrared (Vis/NIR) Hyperspectral Imaging

摘要

In order to identify the growth process of Penicillium, Visible/Near-Infrared (Vis/NIR) hyperspectral imaging with a wavelength range from 400 to 1000 nm was applied. In this paper, the hyperspectral images of the Penicillium growing on the Rose Bengal medium for 6 days were recorded at regular intervals (24h) after inoculation. The average spectral of Penicillium was extracted, and the spectral reflectance value showed a trend of rising first and then decreasing. Then, principal component analysis was performed. With the first three PCs as its input, the support vector machine (SVM) was used to establish the growth stage discrimination model of Penicillium with accuracies of 99.45% and 90.63%. In order to simplify the prediction model, optimal wavelengths were chosen according to loading plots of first three PCs. New optimal wavelengths SVM model was established and the identification accuracies were 100% and 96.83%. Results revealed that it is feasible to identify the growth process of Penicillium based on Vis/NIR hyperspectral imaging.