Research On Nondestructive Determination Of Meat Quality Based On Hyperspectral Imaging Technology

The main research object of this study is beef,Using visible near infrared hyperspectral imaging system(400-1100nm)to analyze the data of 103 samples of beef.After processing the data of multiple pretreatment algorithm and the optimum wavelength selection,the model and forecast model were established for the correction of the fresh degree and water content of beef,And compares the model of evaluation index,In order to find the most suitable method for the rapid nondestructive testing of beef quality,to provide theoretical basis for the on-line detection of internal quality of beef.The main results of the paper are as follows:Firstly,to use a variety of algorithms for preprocessing the obtained beef of the original spectrum data,Then model the multiple linear regression model,compare the model evaluation index,Finally obtains the appropriate pretreatment algorithm of beef freshness and moisture content as a first-order differential width two polynomial S.Golay 9 point convolution derivative method.Then to reduce the dimension of the spectral data after preprocessing,choose the characteristic wavelengths of fresh beef and water content,in order to facilitate subsequent model establishment.Finally,The MLR,BP-ANN and RBF-ANN were used to carry out the calibration set and prediction set model of beef freshness after above processing,the results show that when set the calibration model,The correlation coefficient(R)of RBF-ANN modeling was 0.9983,root mean square error(RMSEC)was 0.0426.when set the prediction model,The correlation coefficient(R)of RBF-ANN modeling was 0.9989,root mean square error(RMSEP)was 0.0256,the best model of the beef freshness is the RBF-ANN model.The MLR,BP-ANN and RBF-ANN were used to carry out the calibration set and prediction set model of beef water content after above processing,the results show that when set the calibration model,The correlation coefficient(R)of RBF-ANN modeling was 0.9989,root mean square error(RMSEC)was 0.0338.when set the prediction model,The correlation coefficient(R)of RBF-ANN modeling was 0.9986,root mean square error(RMSEP)was 0.0306,the best model of the beef water content is the RBF-ANN model.In summary,the effect of RBF-ANN on beef quality modeling was the best.Italso laid the foundation for the rapid and nondestructive testing of beef.