Non-Destructive Evaluation Of Chilled Pork Freshness Based On Fluorescence Spectroscopy Using Excitation-Emission Matrices

China is one of the largest pork production and consumption contries with chiled pork as the dominant form for pork comsumption.For the rapid non-destructive evaluation of pork freshness,fluorescence spectroscopy with the properties of good specificity,highly sensitivity,rapid,non-destructive and relatively low cost,have attracted the attention of researchers.Fluorescence spectroscopy using excitation-emission matrices(EEM)is a new fluorescence spectroscopy technology developed in the past three decades.With this technology,the change of fluorescence intensity combined variation in excitation and emission wavelength can be acquired.The study utilized the chilled pork as the research object.First the raw EEM was analyzed.And with parallel factor analysis,fluorescent components were separated from the raw EEM of chilled pork.Then Characteristic excitation emission matrices were constructed based on fluorescence components.Finally,the research on non-destructive evaluation of chilled pork freshness was studied.The main research contents,results and conclusions were listed as follows:(1)The number of fluorescence peaks,variation of fluorescence peak locations and peak value in the raw EEM of chilled pork was compared.The results showed that:1)There were three peaks.The excitation wavelength(λEx)/emission wavelength(λEm)of A,B,and C in EEM were 250-310 nm/300-400 nm,300-450 nm/400-550 nm,320-450 nm/590-610 nm,respectively.2)The changing trend of the fluorescence peak A value was similar in lean and fat:the fluorescence peak A value decreased rapidly during the first two days,while it decreased slowly during the following 2 days when stored at 20℃.But the fluorescence peak A value decreased more slowly when the samples stored at 4℃,And the changing trend of the fluorescence peak B value was also similar in lean and fat:the fluorescence peak B value increased slowly during the first two days,but increased fairly rapidly during the following 2 days when stored at 20℃.While the fluorescence peak B value increased slowly when the samples stored at 4℃.(2)Fluorescent components separation methods of chilled pork EEM was studied.Additionally,the fluorescence substances were studied by comparison with previous studies.The results displayed that with parallel factor analysis,five main fluorescent components(C1,C2,C3,C4 and C5)were separated from the raw EEM of chilled pork samples,peak positions of λEx/λEm were 260-320 nm/300-400 nm,290-320 nm/300-400 nm,260-440 nm/360-500 nm,390-480 nm/420-550 nm,440-500 nm/470-550 nm,respectively.Combined with the previous literatures,fluorescent components of C1-C4 were identified as tryptophan,vitamin B6,NADH and riboflavin,respectively.But C5(440-500 nm/470-550 nm)remained as unknown.(3)The influence of different classification methods,including support vector machine(SVM)and partial least squares(PLS)combined with different modeling excitation emission matrices were compared.The results illustrated that:1)PLS model produced higher level of classification rate than SVM model using excitation emission matrices with single excitation wavelength based on same fluorescence substances.And the total classification accuracy of SMLR and PLS model of were both the highest.2)PLS model produced the highest classification rate using excitation emission matrices with multiple excitation wavelength based on fluorescence substances of C5(XC5).And total classification accuracy of prediction of PLS model reached to 87.9%.(4)The influence of different quantitative methods,including stepwise multiple linear regression(SVM)and partial least squares regression(PLSR)combined with different excitation emission matrices were used to evaluate the chilled pork freshness.The results demonstrated that:1)The performance of PLSR chilled pork freshness model using excitation emission matrices based on same fluorescence substances,was better than SVM model.2)The performance of PLSR TVB-N model using excitation emission matrices with multiple excitation wavelength based on fluorescence substances of C4(XC4)was optimal,with the correlation coefficient(r)of 0.85 and root mean square error of prediction(RMSEP)of 6.91 mg/100 g.The performance of PLSR TBA model using excitation emission matrices based on all fluorescence substances(Xall)was optimal(r = 0.80,RMSEP = 0.0316 mg/kg).And the performance of PLSR pH model using excitation emission matrices based on Xall was optimal(r = 0.90,RMSEP=0.2789).