Determination of aflatoxin B1 value in corn based on Fourier transform near-infrared spectroscopy: Comparison of optimization effect of characteristic wavelengths
Tao Liu, Jiarong He, Wenzhe Yao, Hui Jiang, Quansheng Chen
Abstract
Aflatoxin B1 (AFB1) is an important cause of human liver cancer. This study proposes a quantitative detection method for the AFB1 in corn based on Fourier transform near-infrared (FT-NIR) spectroscopy technology. First, we acquired spectral data of corn samples with different degrees of mildew using FT-NIR spectrometer; then, we used the ant colony optimization (ACO) and the NSGA-Ⅱ algorithms to optimize the characteristic wavelengths of the spectra after SNV treatment respectively; Finally, the back propagation neural network (BPNN) models were established using the optimized characteristic wavelengths to realize the accurate detection of the AFB1 in corn. The results obtained showed that the prediction performance of the BPNN model established using the four characteristic wavelength variables optimized by NSGA-Ⅱ algorithm is the best, and the best NSGA-Ⅱ-BPNN model's correlation coefficient of prediction (RP) is 0.9951, the root mean square error of prediction (RMSEP) is 1.5606 μg ⋅ kg−1. The overall results demonstrate that the quantitative detection of AFB1 in corn by FT-NIR technique is feasible; in addition, the NSGA-Ⅱ algorithm has its unique advantages in the optimization of spectral characteristics, and it can obtain characteristic wavelength variables with strong pertinence and a small number.