Laser Induced Fluorescence Hyperspectral Imaging (LIF-HSI) for Aflatoxin B1 Detection in Pistachio Kernels

Qifang Wu,Hao Tian,Huirong Xu

DOI: https://doi.org/10.13031/aim.201900103

2019-01-01

Abstract:Pistachio is susceptible to aflatoxins contamination due to its rich nutrients during growth and storage stages. Fluorescence hyperspectral imaging (HSI) technique has been employed recently for screening contaminated agricultural products, nevertheless, several considerable challenges like weak signal intensity and unsatisfactory specificity resulted from common light source with low excitation efficiency and poor monochromaticity, remain unresolved. In this regard, a laser induced fluorescence HSI (LIF-HSI) system (408-1007 nm) was proposed and used for detecting 300 pistachio kernels artificially contaminated with aflatoxin B1 (AFB1, including 5, 10, 20, 30, 50 ppb). The results of principal component analysis (PCA) showed a visual separation between control and contaminated kernels. Generally, the best overall accuracy of ≥ 98.0% was obtained as classifying all contaminated samples by support vector machine (SVM) using the 10 wavelengths extracted by stepwise multiple linear regression analysis (SMLR). The root mean square error of validation (RMSEP) of partial least squares regression (PLSR) model coupled with successive projections algorithm (SPA) using only 11.1% of variables was lower than 5.50 ppb, indicating an interesting potential for discriminating low level of AFB1 contaminated samples. In summary, the LIF-HSI approach had a preliminary feasibility for identifying pistachio kernels contaminated AFB1 qualitatively, and further work should be conducted on the real-time detection study based on naturally contaminated samples for meeting high throughput requirements in real conditions.