| Mutton has always been the main meat product in XinJiang and other northwest plateau areas.In recent years, mutton cates such as the ghanam mashwiy, shish kebab and lamb pilaf have gradually been formed into XinJiang characteristics, which have also gradually been favorited by the national people. In this way, XinJiang mutton market begins to run along the industrial road of the high-ending and branding. However, the mutton quality detection exists a lot of loopholes in the process of transportation and market circulation, which is mainly ascribed to the disadvantages of tedious and time-consuming operations of traditional detection methods incluing sensory evaluation,conventional physical, chemical and microbiological examination, which can’t meet the large quantities, scale and real-time detection on fresh mutton. Thus, the study applied hyperspectral imaging(HSI) technique coupled with chemometric methods to predict and visualize mutton attributes(pH and TVC). The encouraging results can provide a possible way for mutton inspection.The main contents and results in this article were included as follows:(1) Quantitative analysis research was carried on the non-vacuum packaging chilling mutton pH. Firstly, impact study of the two different ROIs methods "image segmentation" and "rectangle region" on mutton pH models were contrast and analyzed in this experiment. It was concluded that:(1) For the three kinds of modeling methods of SMLR, PCR and PLSR, the optimal models built by the spectral information extracted from "image segmentation" and "rectangle region"were PLSR and SMLR, respectively;(2) Comparing with the "rectangle region", models corresponding to"image segmentation” all have a better effect.(3) In conclusion, PLSR regression analysis model established by spectral information extracted from "image segmentation" was the optimal. Its’ LVs is 12, the correlation coefficient(Rc) and root mean square error(RMSEC) of calibration set are 0.95 and 0.050 respectively, the correlation coefficient(Rp) and root mean square error(RMSEP) of prediction set are 0.91 and 0.071, respectively. Secondly, W-PLS, SiPLS-PLS, SiPLS-GA-PLS,UVE-PLS, CARS-PLS and SPA-PLS model effects of of the vacuum packaging chilling mutton pH were compared and analyzed, which concluded that PLSR model effect based on the 28 wavelength points screened by CARS was the best. The correlation coefficient(Rc) and root mean square error(RMSEC) of calibration set are 0.98 and 0.033 respectively, the correlation coefficient(Rp) and root mean square error(RMSEP) of prediction set was 0.96 and 0.053, respectively. This best model was then used to form the distribution map of pH. The results indicated that the CARS-PLS model can be applied to the detection on the non-vacuum packaging chilling mutton pH.(2) Quantitative analysis research was carried on the vacuum packaging chilling mutton pH.Firstly, PLSR model effects of the chilling mutton under different preprocessing methods were contrasted and analyzed, which selected the optimum preprocessing method as the second derivative(2d), S-G smoothing(23 points), multiple scatter correction(MSC) and mean-centering.Secondly, SiPLS, GA, UVE, CARS and SPA were used to extract characteristic bands of chilling mutton spectra respectively, and model effects of W-PLS, SiPLS-PLS, GA-PLS, UVE-PLS,CARS-PLS and SPA-PLS were compared and analyzed, which concluded that the PLSR model result established by 47 wavelength points filted by CARS was best. Its optimal LVs is 13, Rc, Rcv and Rp were 0.98, 0.96 and 0.96, the RMSEC, RMSECV and RMSEP were 0.062, 0.089 and 0.068,and RPD of prediction set was 4.88. The results indicated that CARS-PLS model can predict the vacuum packaging lamb pH. After that, visual distribution map of chilling mutton pH in vacuum packaging was established based on the CARS-PLS regression analysis model.(3) Quantitative analysis research was carried on chilling mutton total viable bacteria(TVC)with the vacuum packaging in the second batch. PLSR model effects of the chilling mutton under different preprocessing methods were contrasted and analyzed, which selected the optimum preprocessing method as the second derivative(2d), S-G smoothing(13 points) and mean-centering.SiPLS, GA, UVE, CARS and SPA were used to extract characteristic bands of chilling mutton spectra respectively, and model effects of W-PLS, SiPLS-PLS, GA-PLS, UVE-PLS, CARS-PLS and SPA-PLS were compared and analyzed, which concluded that the PLSR model result established by 70 wavelength points filted by CARS was best. Its optimal LVs is 13, Rcal and RMSEC of calibration set were 0.98 and 0.29, Rcv and RMSECV of cross-validation were 0.96 and0.46, the Rp and RMSEP of prediction set were 0.96 and 0.47, respectively. And RPD of prediction set was 3.58, which indicated that model have better prediction performance for vacuum packaging lamb TVC. After that, visual distribution map of chilling mutton total viable bateria in vacuum packaging was established based on the CARS-PLS regression analysis model. |
PDF Full Text Request