In this study, visible-near infrared spectroscopy over the spectral range of 400–2500 nm was utilized to detect surface contamination of corn kernels with aflatoxin B1. The artificially contaminated samples were prepared by dropping known amounts of aflatoxin B1 standard dissolved in 50:50 (v/v) methanol/water solution, onto corn kernel surface to achieve different contamination levels of 10, 20, 50, 100, 500, and 1000 ppb. Both endosperm and germ sides of corn kernels were used for artificial contamination, and a total of 210 contaminated and control kernels were scanned with the visible-near infrared spectroscopy in reflectance mode. Spectral preprocessing methods including standard normal variate, first derivative, second derivative, first derivative + standard normal variate, and second derivative + standard normal variate were applied on the original absorbance spectra. Using the original and preprocessed spectra, the 3-class and 7-class discriminant models were established by the chemometric methods of principal component analysis-linear discriminant analysis and partial least squares discriminant analysis separately. The results showed that in discriminating the aflatoxin B1 contamination levels, the spectral range II (1120–2470 nm) generally performed better than using the corresponding spectra type over range I (410–1070 nm). Compared to using the original spectra, the first derivative and second derivative spectra generally improved the performance of the classification models. For classification thresholds of 20 and 100 ppb in aflatoxin B1, the best 3-class models achieved the same overall accuracy of 98.6% for prediction over both ranges I and II. For the 7-class discriminant models, the best overall accuracies obtained over ranges I and II are 91.4 and 97.1% for prediction.
© 2019 The Author(s)PDF Article