Rapid Identification Of Lipid Oxidation In Fresh Chilled Meat By High-resolution Spectroscopy

It is vital and very important to evaluate the freshness of meat quickly and accurately,in order to satisfy the increasing demand of consumers for meat quality and safety.And It is generally known that the degree of fat oxidation is one of the indexes to measure the freshness of pork.However,the current cold meat industry has not been able to detect the degree of fat oxidation quickly and quantitatively in a non-destructive way.Based on high resolution near-infrared spectroscopy at the range of 900¹700nm,this study explored the quantitative prediction of Thiobarbituric acid value,peroxide value and acid value of pork and chicken refrigerated for 0¹4 days at 0⁴℃.So as to achieve the detection of the degree of fat oxidation in chilled fresh pork and chicken non-destructively and rapidly.The main research contents and results are as follows:（1）The thiobarbituric acid（TBA）value of pork stored for 0¹4 days in cold storage at 0⁴℃was rapidly and quantitatively detected by high resolution spectroscopy.Firstly,spectral images of pork samples at different storage periods from 900 nm to 1700 nm were collected,and then the average spectral information of the region of interest of the samples was extracted.Next,the spectral data were preprocessed with Moving Average Smoothing（MAS）,Savitzky-Golay Smoothing（SGS）,Median Filter Smoothing（MFS）,Gaussian Filter Smoothing（GFS）,and Normalization Correction（NC）,Multiplicative Scatter Correction（MSC）,Baseline Correction（BC）and Standard Normal Variable Correction（SNV）.The above spectral data was combined with partial least square regression（PLSR）algorithm to construct a full-band prediction model of pork TBA value.Next,the full-band prediction model of pork TBA values was constructed by partial least square regression（PLSR）algorithm combined with the spectral data mentioned above.By comparing the accuracy and stability of the prediction model,it is determined that GFS is the best spectral pretreatment method.To optimize the whole band model,Regression Coefficient（RC）,Stepwise and Successive Projections Algorithm（SPA）of three algorithms were used to screen the optimal wavelengths and reconstructed based on the optimal wavelengths PLSR model and Multiple Linear Regression model.To optimize the full-band model,Regression Coefficient,Stepwise Regression（Stepwise）and Successive Projections Algorithm（SPA）three algorithms were used to screen the optimal wavelengths and reconstructed the PLSR model and Multiple Linear Regression（MLR）model.The results showed that the prediction accuracy of GFS-RC-MLR model and GFS-RC-PLSR model according to the29 optimal wavelengths extracted by RC algorithm were similar and satisfactory,with the regression coefficient of prediction（r _P）0.924,and the root mean square error of prediction（RMSEP）0.035mg/100g and 0.034mg/100g,respectively.（2）The feasibility analysis of quantitative detection of TBA value of chicken meat refrigerated for 0⁷days at 0⁴℃by long wave near-infrared hyperspectral imaging.The spectral information in the hyperspectral images of chicken samples with different degrees of fat oxidation were obtained and the pretreatment of MAS,SGS,MFS,GFS,NC,MSC,BC and SNV were utilized in the experiment.Then,three methods of RC,Stepwise and SPA were applied to reduce the dimension of the huge spectral information,and the optimal wavelength was selected respectively.The results shows that better characterization of chicken TBA values of 31 wavelength is elected by RC algorithm.And,the GFS-RC-PLSR model（r_P=0.945,RMSEP=0.053mg/100g）was better than the GFS-RC-MLR model（r_P=0.934,RMSEP=0.058mg/100g）based on the optimal wavelength.The prediction accuracy of both models was ideal,indicating that it was feasible to evaluate the degree of chicken fat oxidation according to the hyperspectral technology.（3）The peroxide value was used as the evaluation index of the degree of oxidation of pork fat to study the freshness of pork by hyperspectral imaging analysis.Firstly,the near-infrared spectral data of pork samples collected in the experiment were preprocessed to construct the prediction model of full-band PLSR of RAW spectrum and preprocessed spectrum respectively.The results show that GFS is the best pretreatment method.Then,three algorithms,RC,Stepwise and SPA,were used to reduce the dimension of GFS spectrum,and the optimal wavelength was extracted and the optimization model was established.The results showed that the GFS-SPA-MLR model with pork peroxide value(r_P=0.877,RMSEP=7.55×10^-4g/100g)had better prediction effect on the 25 optimal wavelengths screened by SPA algorithm.（4）The acid value of pork was determined by hyperspectral technique.First of all,the average spectrum was extracted from the hyperspectral image of the collected pork samples,and then the RAW spectrum was preprocessed by a variety of methods（MAS,SGS,MFS,GFS,NC,MSC,BC and SNV）.By comparison,it is found that the model with RAW and BC spectra has better prediction effect.Finally,RC and SPA algorithms were used to select the optimal wavelength and rebuild the PLSR and MLR prediction models,respectively.The results showed that the RAW-RC-PLSR model（r_P=0.846,RMSEP=0.569 mg/g）with 28 optimal wavelengths extracted by RC algorithm had higher correlation coefficient,smaller error and better prediction performance.