Detection Of Urea Additives In Milk Based On Hyperspectral Imaging Technology

As a common nutrition in people’s daily life,milk plays an important role in maintaining human health.In the market,there are certain food additives in milk to improve the nutritional value of milk.However,unscrupulous manufacturers often mix harmful additives such as urea into milk,so as to seek greater benefits.Excessive urea content poses a great threat to seriously endangers the health of people,which even has the risk of death.In this paper,milk mixed with urea was taken as the research object,hyperspectral data of milk mixed with urea were obtained by using hyperspectral imaging technology,spectral preprocessing algorithm,wavelength selection and model optimization method,and the analysis model was established.Two long short-term memory network models(LSTM)based on bionic optimization algorithm were proposed,so as to realize the(accurate)regression prediction of urea content in milk,and achieve an integrated prediction system for milk urea additives.The specific work is as follows:1.Two methods to optimize the network structure of LSTM model by using whale optimization algorithm(WOA)and grey wolf optimizer(GWO)was proposed.Taking partial least squares regression as the baseline model,the pretreatment method and characteristic band were determined,and a variety of regression models were established.The LSTM model with the best prediction effect was selected for optimization.Compared with the original LSTM model,the test set determination coefficient R_p~2=0.9882 and the root mean square error RMSEP=0.5600 of the model optimized by WOA were improved by0.32%and 1.65%,respectively.Compared with the original LSTM model,the decision coefficients R_p~2=0.9906 and RMSEP=0.4996 of the model optimized by GWO were improved by 0.56%and 7.69%,respectively.Experiments show that the two optimized models can accurately predict the urea content in milk,which meets the requirements of experimental model accuracy and realizes the accurate prediction of urea content in milk.2.The integrated prediction system of milk urea additive was Achieved,which mainly has the functions of hyperspectral data preprocessing,band screening and regression model establishment.5 pretreatment methods,4 wavelength screening methods and 3 regression models were provided for users to use.Among them,regression models include machine learning,ensemble learning and deep learning models.The experimental results showed that hyperspectral imaging technology combined with deep learning method could effectively detect urea content in milk,which provides a theoretical basis and method for the detection of urea content in milk.