Determination Of Phosphate And Potassium Sorbate Content In Hairtail And Cod Surimi And Surimi-based Seafood By Multi-fingerprints Technology

In order to achieve rapid nondestructive determination of phosphate andpotassium sorbate content in surimi and surimi-based seafood, in this study weresearched the application of near infrared spectroscopy (NIRS) technology andelectronic nose technology in the quantitative rapid detection of the phosphate andpotassium sorbate content in surimi and surimi-based seafood. We want to providetechnical support for rapid detection of the quality of surimi and surimi-based seafood.The results are as follows:(1) NIRS calibration models of phosphate in hairtail surimi and cod surimi wereestablished by partial least squares regression (PLS). The result shows that correlationcoefficients calibration in the model of hairtail surimi and cod surimi is0.983and0.968, respectively, and the prediction of calibration is0.032and0.045, respectively.The correlation coefficients of the validation are0.951and0.966, respectively. Theprediction of validation is0.058and0.048, respectively. When the confidence levelwere95%, no significant difference was found between the NIRS analysis andreference analysis (P＞0.5). Two NIRS models were unified, and the correlationcoefficients of calibration and validation is0.947. The prediction of validation andcalibration is0.079and0.078, respectively. The PLS models were compared withsupport vector regression (SVR) models, and the PLS models prediction results werebetter than SVR models.(2) NIRS calibration models of phosphate in hairtail kamboko and cod fish ballwere established by partial least squares regression. The result shows that correlationcoefficients calibration in the model of kambokoo and fish ball is0.960and0.946,respectively, and prediction of calibration is0.101and0.058, respectively. Thecorrelation coefficients of the validation are0.954and0.947, respectively. The prediction of validation is0.058and0.042, respectively. When the confidence levelwere95%, no significant difference was found between the NIRS anlysis andreference analysis (P＞0.5). Two NIRS models were unified, the correlationcoefficients of calibration and validation is0.904and0.894. The effect is not as goodas the separate modeling.(3) As the output for the predicted content of phosphate by the electronic nose insurimi was fitted with original property by PLS. The linear of fitted curve is good andthe correlation coefficient of hairtail surimi and cod surimi is0.936and0.941,respectively. The PLS models were compared with SVR models, the PLS modelsprediction results were better than that of SVR models.As outputting the predictedcontent of phosphate by the electronic nose in hairtail kamboko and cod fish ball werefitted with original property by PLS. The linear of fitted curve is good and thecorrelation coefficient of hairtail surimi and cod surimi is0.918and0.924,respectively. The study demonstrated that the electronic nose could be used to detectthe contents of phosphate in surimi-based food (kamaboko and fishball) to someextent.(4) Different pretreatment methods were compared and first-order derivatives gap2(db1g2) and normalization by closure (ncl) was applied with the best spectral range5000~7144,7404~10000. The PLS was used fitted the original property with predictedproperty of potassium sorbate content in kamboko. The result shows that coefficientscorrelation in the calibration and validation of the model of kamboko is0.966and0.960, respectively. The prediction of calibration and validation is0.032and0.037,respectively.(5) The fish ball was researched. The PLS models were compared with SVRmodels, the correlation coefficients of SVR model and PLS model is0.975and0.923,respectively. So the SVR model prediction results were better than PLS model.(6) The content of potassium sorbate in kambokoo and fish ball were studied bye-nose. The injected volume was optimized by E-nose. The optimal parameters:injection sample2.0g, injection volume1500μL, and headspace equilibration time 600s. As outputting the predicted content of potassium sorbate by the electronic nosein kamboko was fitted with original property by PLS. The correlation coefficient ofhairtail kamaboko is0.916. The PLS models were compared with SVR models, andthe correlation coefficients of SVR model is0.952, so the SVR model predictionresults were better than that of PLS model. The study demonstrated that the electronicnose could be used to detect the content of potassium sorbate in surimi-based food(kamaboko) to some extent. As outputting the predicted content of potassium sorbateby the electronic nose in fish ball was fitted with original property by PLS. Thecorrelation coefficient of fish ball is0.928. The study demonstrated that the electronicnose could be used to detect the content of potassium sorbate in surimi-based food(fish ball) to some extent.