Study On The Determination Of Characteristic Fatty Acids In Tan Sheep Meat By Near Infrared Spectroscopy

In this study,near-infrared reflectance spectroscopy(NIRS)and gas chromatography were used to construct a Partial least squares regression(PLSR)model of 5 characteristic fatty acids in Tan sheep meat.Different spectral preprocessing methods and interval random leapfrog algorithm were used in the modeling process Random frog(IRF)and competitive adaptive reweighting algorithm(cars)are used to improve the performance of the model.In the experiment,the PLSR model of the characteristic fatty acid content of freeze-dried Tan sheep meat was established by using freeze-drying technology,and compared with the fresh Tan sheep meat model.The specific contents and conclusions are as follows:(1)Establishment of near infrared model of oleic acid and linoleic acid in Tan sheep meat:for the prediction model of oleic acid content,the model with better performance is based on the full wavelength model,and the processing method is SNV+first derivative,Rc=0.8895,RMSECV=10.2515,Rp=0.7315,RMSEP=9.9815;for the prediction model of linoleic acid content,the best performance of the model is after extracting the characteristic wavelength by IRF algorithm,The correlation coefficient(Rc)of the calibration set based on the original spectrum and the spectrum smoothed by S-G can reach more than 0.9.The model Rc of the original spectrum is higher,which is 0.9912,RMSECV is 0.0118,Rp is 0.9879,RMSEP is 0.0122.The results showed that near infrared spectroscopy technology can realize the rough prediction of oleic acid content in Tan sheep meat,and near infrared spectroscopy technology can accurately predict the content of linoleic acid in Tan sheep meat.(2)Establishment of the near-infrared model of palmitic acid,linolenic acid and arachidonic acid in Tan sheep meat:The highest Rc in Tan sheep meat linolenic acid content prediction model is the spectrum model that uses the full wavelength and is processed by MSC and first derivative,which is 0.8414,but the correlation coefficient Rp of its prediction set is 0.5870;for the prediction model of arachidonic acid content,the model based on CARS algorithm has better model performance.Its Rc is 0.8035 and Rp is 0.7560,which is compared with other established peanuts.The tetraenoic acid content model has a certain improvement in Rp;the established palmitic acid content prediction model has poor performance.The results show that the rapid determination model of these three fatty acids by near-infrared spectroscopy needs to be further improved.(3)Comparison of the accuracy of the near-infrared model of five characteristic fatty acids in fresh Tan sheep meat and freeze-dried Tan sheep meat:The oleic acid content model in freeze-dried Tan sheep meat is better based on the IRF algorithm.After MSC+first-order derivation processing,Rc reaches 0.9531,RMSECV is 11.1913,corresponding Rp is 0.8076,RMSEP is 8.3919.The prediction model of palmitic acid content in freeze-dried Tan sheep meat has better performance based on the IRF algorithm.The pretreatment methods with higher Rc are SNV+first-order derivative,MSC+first-order derivative and first-order derivative,respectively 0.8318,0.8315 and 0.8325,the corresponding Rp is 0.8302,0.8284,0.8135,RMSEP is 9.2141,9.2258,and 9.1928,and the correlation coefficients of the prediction set are all above 0.8;The freeze-dried Tan sheep meat model established based on the CARS algorithm has Rc of 0.8667 and RMSECV of 9.4860.Rp is 0.8339 and RMSEP is 8.2375.After freeze-drying,the linolenic acid content prediction model Rc established based on the IRF algorithm has been improved,but it is still lower than 0.8.The results show that freeze-drying technology can improve the model performance of oleic acid,palmitic acid and linolenic acid content to a certain extent,but significantly reduce the performance of prediction model for linoleic acid and arachidonic acid content.