Study On Hyperspectral Detection Of Nitrite Change In Sausage During Storage

The paper used hyperspectral techniques to predict the nitrite content of sausages in seven different periods.The results not only improve the accuracy of hyperspectral technology in detecting the nitrite content of sausage during storage,but also provide a new means for quantitative detection of nitrite content in sausage.The main research work is as follows:First,select seven different period of sausage samples to test the nitrite content according to the national standard method.For each period of the sample,40 samples were taken,a total of 280 samples,and the spectral data was collected for each period of the sample by hyperspectral technique,and black and white correction was performed.In order to reduce the interference of external factors on the spectral information,the original spectral data was preprocessed.After comparing the processing results of two kinds of preprocessing methods,multi-scatter correction(MSC)and Savitzky-Golay convolution smoothing(SG smoothing),the selected SG smoothing method is used to preprocess the original spectral data.The acquired hyperspectral information has a total of 1288 bands of spectral data,which will increase the amount of calculation in the modeling process.In order to reduce the complexity of modeling,the spectral wavelengths were extracted by the least squares regression coefficient,and 29 characteristic wavelengths were extracted.Quantitative model of nitrite content in sausages at characteristic wavelengths was constructed by multiple regression,principal component regression and partial least squares regression.The results showed that multiple regression,principal component regression and partial least squares regression model were used to determine the prediction set.The coefficients R~2 are up to 0.8588,0.8961,and 0.9111,respectively,and the corresponding root mean square errors RMSE are 0.1687,0.1488,and 0.1397,respectively.This indicates that the prediction model established at the characteristic wavelength is not ideal and the prediction accuracy is not significantly improved.Based on the extracted characteristic wavelengths,the original data information can not be fully reflected.The paper uses principal component regression and partial least squares regression to establish prediction models directly at full wavelength.In order to compare with the model prediction results constructed under the characteristic wavelength,the first 29 principal components are selected as the input vectors of the regression model.The decision coefficients R~2 of the prediction set of principal component regression and partial least squares regression model at full wavelength are0.9522 and 0.9829,respectively,and the corresponding root mean square errors are0.0974 and 0.0592,respectively.The results show that the prediction model established at full wavelength has higher precision and the prediction result is ideal.When using the hyperspectral technique to quantitatively analyze the nitrite content in sausages,the prediction results at full wavelength are better than those predicted at the characteristic wavelength.It is indicated that full-wavelength analysis is appropriate when using Principal Component Regression and Partial Least Squares Regression to process data on nitrite content in sausages.