DETECTION OF FLAXSEED OIL ADULTERATION BASED ON TWO-DIMENSIONAL CORRELATION NEAR-INFRARED SPECTRA

Abstract

Flaxseed oil is rich in α-linolenic acid and other nutrients, and the adulteration happens frequently because of its high price. To detect the adulteration of flaxseed oil quickly and accurately, a method was proposed based on weighted reconstructed two-dimensional correlation near-infrared(NIR) spectra. The near-infrared spectra of 79 adulterated flaxseed oil samples (adulterated by rapeseed oil with the doping volume ratio 1%-40%) were measured, and the traditional two-dimensional correlation synchronous spectra were calculated. The two-dimensional correlation synchronous spectra of all samples were decomposed into multiple components of different scales by the bi-dimensional empirical mode decomposition algorithm (BEMD). According to the root mean square error(RMSE) values of the adulteration detection sub-models established by each component, the weights of the corresponding components were calculated, and then the two-dimensional correlation spectra of all samples were reconstructed by accumulating the weighted components. A quantitative analysis model of flaxseed oil adulteration was established based on the weighted reconstructed two-dimensional correlation spectra combined with the N-way partial least square(N-PLS)algorithm. Compared with the traditional two-dimensional correlation spectroscopy, the model built by the weighted reconstructed two-dimensional correlation spectra had better performance with the calibration determination coefficient increased by 6.05%, and the prediction determination coefficient increased by 7.5%. The proposed method could effectively enhance the spectral feature information, reduce the spectral noise interference, and hence provide a new idea for the detection of edible oil adulteration.