Study On Detection Of Imported Fresh And Frozen-Thawed Beef Quality Based On Spectroscopy And Imaging Information

China is now the third largest beef producer in the world,and even so,it still imports plenty of beef(mainly fresh and frozen beef)to solve the increasing gap between supply and demand.Recently,take into consideration high market value of fresh and frozen beef,some illegal businessmen try to make improper profits by pretending fresh beef with sub-fresh beef or pretending frozen-thawed beef.Some conventional detection methods has the advantages of high sensitivity and accuracy,however,there are also some obvious disadvantages such as complicated pretreatment and long detection time.With the maturity of near infrared spectroscopy(NIR)and hyperspectral imaging(HSI)technology,they have been widely applied in food safety detection.Therefore,this study attempts to identify fresh beef during different storage times,fresh and frozen-thawed beef and monitor the quality changes of beef during storage using NIR and HSI techniques.The specific content and related conclusions of the research are as follows:(1)Screening of characteristic indicators between fresh beef during different storage times,fresh and frozen-thawed beef.The total volatile basic nitrogen(TVB-N),pH,color parameters(L*,a*,b*)and the content of deoxymyoglobin(DMb),oxygenated myoglobin(OMb)and metmyoglobin(MMb)of samples were detected separately.Based on correlation and significant difference analysis,TVB-N,pH,a*and b*were selected as the characteristic indicators of freshness in fresh beef with different storage times.At the same time,a*,b*and MMb were selected as the differential characteristics indicators of fresh and frozen-thawed beef.(2)Identification of fresh beef during different storage times,fresh and frozen-thawed beef and their storage time.Firstly,spectral and image data were collected by NIR and HSI respectively.And then,different pretreatment methods were adopted to deal with NIR spectral data,HSI spectral data,HSI texture variables and fusion of all HSI data.Afterwards,feature variable screening methods including competitive adaptive reweighted sampling(CARS),interval partial least squares(iPLS),interval partial least squares-competitive adaptive reweighted sampling(iCARS)to establish the identification models including the linear discriminant analysis(LDA),K-nearest neighbors(KNN),back-propagation artificial neural network(BPANN)and random forest(RF).The results showed that fresh beef during different storage times can be identified by above technologies,and HSI technology has the ability to identify fresh and frozen-thawed beef and their storage time.The LDA model constructed based on the spectral data obtained by HSI technology had the optimal recognition effect on identifying fresh beef during different storage times,and the recognition rates of test set was 100%.The RF model constructed based on the spectral data obtained by HSI technology had the optimal recognition effect on identifying fresh and frozen-thawed beef and their storage time,and the recognition rates of test set was 97.86%.(3)Rapid prediction model and visualization distribution maps of the content of characteristic indicators were established using NIR and HSI techniques.Firstly,collecting the NIR spectral data and hyperspectral imaging of samples,and then the region of interest were selected from image to extract mean spectrum.Finally,using NIR spectral data and HSI spectral data with different preprocessing methods,feature variable screening methods including CARS,iPLS and iCARS were used to establish the partial least squares(PLS)prediction model.The results showed that the two techniques can predict the content of characteristic indicators,and the best prediction models were CARS-PLS.The predicted results of HSI for TVB-N and b*of freshness characteristic indicators were better,and the correlation coefficients of prediction set(r_p)were 0.9637 and 0.9423,and the root mean square error of prediction(RMSEP)was 1.12 mg/100g and 0.77,respectively.The predicted results of NIR for pH and a*were better with that r_p were 0.9512,0.9683 and RMSEP were 0.0159,0.59,respectively.HSI was better in predicting the content of differential characteristics indicators of a*,b*and MMb with r_p were 0.9203,0.8506,0.9244 and RMSEP were0.85,1.01,1.73%,respectively.At the same time,using the CARS-PLS model based on HSI spectral data combined with HSI image information to visualize the distribution of each characteristic indicators’content in beef image.The study proved that both NIR and HSI techniques can identify fresh beef at different storage times and predict the content of characteristic indicators.Meanwhile,HSI technology can be used to identify fresh and frozen-thawed beef and their storage time,and also to visualize the distribution of characteristic indicators’content.This study provides an effective theoretical basis for monitoring meat adulteration and standardizing the order of imported beef market.