Variation analysis in spectral indices of volatile chlorpyrifos and non-volatile imidacloprid in jujube (Ziziphus jujuba Mill.) using near-infrared hyperspectral imaging (NIR-HSI) and gas chromatograph-mass spectrometry (GC-MS)

摘要

Two pesticides in terms of chlorpyrifos and imidacloprid contaminating edible jujube fruits were determined using hyperspectral imaging (900-1700 nm) and gas chromatograph-mass spectrometry (GC MS). Hyperspectral images of jujube samples contaminated by pesticides at different concentrations were collected. Their spectral data extracted in reflectance (R-S), absorbance (A(S)), exponent (E-S) and Kubelka-Munck (K-M-S), were respectively used to develop partial least squares discriminant analysis (PLSDA) and locally weighted partial least square regression (LWPLSR) models. Based on these spectral parameters, corresponding models defined as A(S)-PLSDA and E-S-PLSDA acquired optimal results, with correlation coefficients of cross-validation (R-CV) of more than 0.900 for recognition of chlorpyrifos concentrations and R-CV of over 0.713 for identification of concentrations of the imidacloprid. The E-S-LWPLSR model obtained the best R-CV of 0.864 for quantitative determination of chlorpyrifos residuals, and the best R-CV of 0.885 for determination of imidacloprid residuals. The feature wavelengths were selected based on the automatic weighted least squares and gap segment derivative (AWLS-GSD) coupled with regression coefficient (RC) method. The better performance was obtained by the resulting simplified E-S-AWLS-GSD-RC-LWPLSR model established using only eight characteristic wavelengths, with R-CV of 0.757, RMSECV of 3.75 x 10(-3) for chlorpyrifos residuals, and R-CV of 0.898, RMSECV of 0.311 x 10(-3) for imidacloprid residuals. To summarize, hyperspectral imaging technology shows a great potential to predict pesticide residuals of jujube fruit. (C) 2017 Elsevier B.V. All rights reserved.