Research On The Detection Of Beef Quality Based On Hyperspectral Imaging Technology

The detection of beef quality by traditional methods will not only affect the measurement results of physical and chemical indicators,but also cause pollution to the samples,and it is also destructive when detecting certain scalar indicators.Therefore,accurate,non-destructive,rapid detection and evaluation is very important,and also has great significance for promoting the development of meat products enterprises.In recent years,the application of hyperspectral imaging in the detection and evaluation of meat quality has become a hot topic,the hyperspectral imaging technology has been widely used in the safety and quality testing of agricultural products,and also achieved good research results.This paper mainly takes chilled beef as the research object,from the sensory and intrinsic quality parameters of meat quality assessment,combined with hyperspectral imaging technology and chemometric analysis method,the sample is studied and analyzed.The following studies were carried out specifically:Using hyperspectral imaging system to obtain hyperspectral data images of 40 beef samples and the spectral spectrum of the region of interest of the sample was extracted,the tenderness of the beef was characterized by the shear force value.Used the whole band original spectrum,the spectrum preprocessed by standard normal variate transformation(SNV)and the shear force values to establish partial least squares regression(PLSR)model of beef tenderness respectively then evaluate the models.The PLSR model processed by the standard normal variate transformation(SNV)has higher prediction effect,the prediction correlation coefficient(Rp)was 0.823,the prediction root mean square error(RMSEP)was 1.494.Using hyperspectral imaging system to obtain hyperspectral data images and spectral reflectance curve of the region of interest of 120 beef samples in four times,the test interval was one day in the meanwhile the scalar beef pH value measured in accordance with the national standard.The PLSR model of chilled beef pH was established by using the original spectrum of the whole band and the spectral information and corresponding scalar values after MSC,SNV,mean centering,normalization,Savitzky-Golay confluence smoothing pretreatment.By comparison the PLSR model processed by the mean centering has a higher prediction effect,the prediction correlation coefficient(Rp)was 0.7801,the prediction root mean square error(RMSEP)was 0.1153.Optimizing the variables of the original model,combined with the genetic algorithm(GA)can reduce the number of variables in the model from 472 to 62 and achieve the effect of variable simplification,the prediction correlation coefficient(Rp)was improved.Using hyperspectral imaging system to carry out the grade prediction of beef marbling.Obtained 75 beef samples of three grades,collected three-dimensional data blocks of hyperspectral image of steak samples,derived RGB images,segmented images of effective eye muscles,selected ROI from them,then extracted textures in the region based on gray level co-occurrence matrix method.Characteristic parameters and the establishment of multiple linear regression models to predict the beef marbling grades.The prediction decision coefficient(R2)was 0.81,the root mean square error(RMSE)was 0.36,the classification accuracy was 86.7%.