Abstract:

This study used a hyperspectral imaging system (HIS) to measure the spectral response of fungi inoculated
on potato dextrose agar plates. In this work, three methods for calculating HIS parameters, including the mean of whole
spectral response values covering the range of 400–1 000 nm, the spectral response value of the wave peak at 716 nm, and
the score of the first principal component in the whole spectral range of 400–1 000 nm using principal component analysis
(PCA), were used to simulate the growth of fungi. The results showed that the coefficients of determination (R2) of the
simulation models for test datasets of three fungi, Botrytis cinerea, Rhizopus stolonifer and Colletotrichum acutatum, were
0.722 3–0.991 4, and the sum square error (SSE) and root mean square error (RMSE) were in a range of 2.03 × 10-4–5.34 × 10-3
and 0.011–0.756, respectively, based on the three methods. The correlation coefficients between HIS parameters and colony
forming units of fungi were high ranging from 0.887 to 0.957. In addition, fungal species can be discriminated by PCA and
partial least squares-discrimination analysis (PLS-DA) based on the spectral information in the full wavelength range. The
classification accuracy of the test dataset by PLS-DA models for fungi cultured for 36 h was 97.5% among Botrytis cinerea,
Rhizopus stolonifer, Colletotrichum acutatum, and the control. This paper offers a new technique and useful information for
further study into modeling the growth of fungi and detecting fruit spoilage caused by fungi based on HIS.

Key words: spoilage fungi, hyperspectral imaging, growth simulation, partial least squares-discriminant analysis (PLS-DA);discrimination