Study On Different Varieties Of Beef Recognition Method Based On Hyperspectral Technology Research And Analysis Of Differences In Quality

In this paper,beef cattle(Angus,Limousin,Simmental,Qinchuan)and dairy cows were selected as the research objects.The physical and chemical indexes of neck meat,bovine meat,gourd meat and tenderloin of five kinds of beef were detected respectively.At the same time,using Vis-NIR hyperspectral imaging technology and NIR hyperspectral imaging technology,five kinds of beef identification models were established by using partial least squares discriminant analysis(PLS-DA),linear discriminant analysis(LDA)and support vector machine(SVM).The main results are as follows:(1)There were differences in physical and chemical indexes among different beef varieties:Compared with beef cattle,the color of cow meat was poor,and the color difference was very significant compared with beef cattle(P<0.01),water content and protein content were lower,pH value,shear force were higher,and quality was poor.Compared with the beef cattle,Qinchuan and Limousin had higher moisture contents,which were 75.33%and 75.98%.And the protein contents were also higher,which were 22.47%and 22.07%respectively.The minimum shear force of Qinchuan cattle was 5.96 kg.The qualities of Angus and Simmental were not very different.(2)There were significant differences in the physical and chemical indicators of the four parts(neck meat,bovine meat,gourd meat and tenderloin)of Limousin and Qinchuan cattle.The quality of neck meat was poor,and the quality of tenderloin meat was better.(3)Six kinds of spectral preprocessing were performed on the original spectra of 400?1000 nm and 900?1700 nm.The optimal pretreatment methods for two bands were selected for multivariate scattering correction and convolution smoothing.(4)There were obvious differences between the average spectral curves of beef of different breeds at 400?1000 nm and 900?1700 nm,and the differences between beef cattle and beef were obvious.The reflectivity of Qinchuan and Limousin neck meat was lower,the tenderloin was higher.(5)The LDA method was superior to SVM and PLS-DA in the 400?1000 nm band,and the Mahalanobis distance method in LDA was preferred,the recognition rate of the correction set was 96.62%,and the recognition rate of the prediction set was 87.23%.The Mahalanobis distance method in LDA was best in the 900?1700 nm band,the recognition rate of the correction set was 95.14%,and the recognition rate of the prediction set was 83.33%.The correct recognition rate of dairy beef was higher than that of the four beef cattle.(6)Comparing and analyzing the LDA models of extracting characteristic wavelengths of beta coefficient,UVE and SPA,the best results of the 400?1000 nm and 900?1700 nm two wavelengths were UVE and SPA.In the wavelength band of 400?1000 nm,23 characteristic wavelengths were extracted by UVE.The recognition rates of the correction set and prediction set were 97.30%and 81.56%.In the wavelength band of 900?1700 nm,10 characteristic wavelengths were extracted by SPA.The recognition rates of the correction set and prediction set were 91.39%and 80.21%.The methods of UVE and SPA can promote the qualitative analysis of different beef varieties and provide technical support for the development of beef cattle industry.