Study On Multispectral Fusion Technology For Detection Of Flour Authenticity

Flour represents one of the most important daily diets,its authenticity issue has received much attention.The routine flour authenticity detection often relies on mass spectrometry and chromatography.However,these methods have the disadvantages of cumbersome steps,expensive instruments,and long detection period,making it difficult to screen hundreds of millions of tons of flour efficiently.It is thus extremely important to develop new and efficient methods for detection of flour authenticity.This paper aims to combine Raman spectrum and laser-induced breakdown spectrum(LIBS)by multispectral fusion technology(MFT).The MFT full use of the complementary properties between the molecular information in the Raman spectrum and the atomic information in the laser-induced breakdown spectrum,which is capable of obtaining more characteristics information of the flour system.In the sequence,chemometrics methods are developed to encode the multispectral data for detection of flour authenticity.With the spectrum characteristics of Raman and LIBS in complex flour systems at hand,this paper develops corresponding spectral preprocessing algorithms and data fusion algorithms.In the process of spectrum pretreatment,the spectrum was decomposed by discrete wavelet transform and adaptive wavelet transform to remove background and noise information form Raman spectra and LIBS spectra,respectively.Thereafter,the competitive adaptive reweighted sampling algorithm was used for variable selection,which removed redundant information from the multispectral data.After signal pretreatment,the data fusion strategy was performed at feature level and decision level,respectively.As a result,qualitative and quantitative analysis of the flour authenticity is achieved.Multispectral fusion models of four common flour doping systems are established to validate the performance of the MFT.The calculation results show that the MFT can not only improve the qualitative discrimination accuracy of the prediction model,but also improve the quantification accuracy of the prediction model.The results demonstrate the feasibility and effectiveness of MFT in the detection of flour authenticity,which may well extend to screen authenticity of other foods as well.