Research On Detection Technology Of Talc Content In Flour Based On NIRS Technology

Flour is one of the main raw materials for the daily diet,and its quality has always been the focus of public attention.In all evaluation indicators of flour,appearance and taste play the key roles in the quality certification of flour and market price.In order to some flour producers add a large amount of talc to the flour to enhance the color and physical properties of the flour.However,many studies have confirmed that talc could actually cause harm to human health.Therefore,it is of great practical significance to study the detection method of talc in flour for the protection of consumers’ health.Nearinfrared spectroscopy technology has been widely used in many fields because of its fast,high-efficiency,non-destructive technical characteristics and advantages,the nearinfrared spectroscopy technology will be used for the quantitative detection of talc in flour in the research.In this study,41 flour samples with different contents of talc were prepared,and the diffuse reflectance spectra of all the samples in the spectrum range of 400 ~ 2500 nm were collected by NIRS analyzer for spectral analysis.Since the selection of training samples is very important in the process of establishing the near-infrared spectroscopy model,two abnormal samples were removed by Mahalanobis distance method,and 29 samples in the normal 39 flour samples were selected for modeling,and the remaining 10 samples were used for prediction as the result of SPXY division.In addition,in order to remove the influence of noise interference on the spectrum signal,different preprocessing methods such as smoothing(SG),multivariate scatter correction(MSC),derivative algorithm(1D,2D)and normalized normal variation(SNV)were compared,the SNV showed the best calibration effect.The correlation coefficient method(CCM)and successive projection algorithm(SPA)and combination of the two methods were used to extract the characteristic wavelength to achieve spectral dimension reduction,and the extracted characteristic spectral data were respectively input into three quantitative analysis models-the partial least squares(PLS)and radial basis function(RBF)artificial neural networks and back propagation(BP)artificial neural network model for quantitative analysis for prediction of talc content in flour.Finally,the model named CCM-SPA-RBF was determined as the best predictive model in the research,with a determination of Prediction Coefficients of Prediction(Rp)of 0.9992,Root-Mean-Square Error of Predication(RMSEP)of 0.1638 and Residual Predictive Deviation(RPD)of 25.2485,respectively.The results showed that the near-infrared spectroscopy combined with multivariate analysis methods can be used to detect the content of talc in wheat flour in a rapid and non-destructive way,which was of positive significance to the safety of wheat flour.