Accurate Discrimination of Mold-Damaged Using Partial Least-Squares Discriminant Analysis and Selected Wavelengths.

Unscrupulous merchants sell the mold-damaged (CRP) after removing the mold. In this study, an accurate and non-destructive strategy was developed for the discrimination of mold-damaged CRPs using portable near-infrared (NIR) spectroscopy and chemometrics. The outer surface and inner surface spectra were obtained without destroying CRPs. The discrimination models were established using partial least squares-discriminant analysis (PLS-DA) and wavelength selection strategy was used to further improve the discrimination ability. The predictive ability of models was assessed using the test set and an independent test set obtained one month later. The results demonstrate that the models of the outer surface outperform those of the inner surface. With multiplicative scatter correction (MSC)-PLS-DA, 100% accuracies were obtained in test and independent test sets. Furthermore, the wavelength selection strategy simplified the models with 100% discrimination accuracy. In addition, the randomization test (RT)-PLS-DA model developed in this study combines both the benefits of high accuracy and robustness, which can be applied for the accurate discrimination of mold-damaged CRPs.