Rapid Detection For Water-holding And Texture Profile Characteristics Of Tan-Mutton Based On Hyperspectral Technology

Water-holding quality and texture profile characteristics are important sensory qualities which can affect the edible value and commercial value of Tan-mutton.Traditional sensory evaluation and physical and chemical testing are difficult to meet the needs of fast online detection.In addition,single sensor non-destructive testing technology is difficult to evaluate meat quality comprehensively and accurately.Therefore,the rapid detection,accurate prediction and safe storage time identification of Tan-mutton are current problems,which needed to be solved urgently.In this paper,the samples of Yanchi Tan-mutton were used as the research subject.The physical and chemical indicators were used to analyze the mechanism of water holding capacity and texture profile characteristics of of Tan-mutton during chilled storage.The hyperspectral image information of Tan-mutton samples was collected by hyperspectral imaging system,and the reflectance spectrum curve of the region of interest in Tan-mutton samples was extracted by the image processing technique.the prediction model of water holding capacities and texture profile characteristics of chilled Tan-mutton and the storage time discriminating model of Tan-mutton were established by the parameter optimization of computer programming technology,which would achieve the rapid,non-destructive,real-time detection of meat and meat products.The main research results are as follows:(1)Hyperspectral imaging technology predicted the water-holding quality and texture profile characteristics of Tan-muttonThe VIS/NIR hyperspectral system was used to acquire hyperspectral image information in the wavelength range of 400～1000 nm,and the spectral information on the image was extracted by ENVI 4.5 software.The partial-least-squares regression(PLSR)algorithm combined with stoichiometry were used to establish a full-band prediction model for different spectral preprocessings of MC and SNV,and the best combination of MC+MC and SNV+MC showed good results.The VCPA and SPA characteristic wavelength extraction algorithms were used to reduce the spectrum,and the PLSR and LSSVM quantitative analysis models for the water-holding quality and texture properties of chilled Tan-mutton were established.The results showed that the VCPA-LSSVM algorithm was effective in predicting the water-holding quality and texture profile characteristics of Tan-mutton,the performance of juice loss,cooking loss,hardness,adhesiveness,resilience,cohesion,elasticity,cohesiveness,chewability models were evaluated by Rc and Rp,which were:0.758,0.761,0.783,0.625,0.572,0.720,0.864,0.872,0.982 and 0.725,0.688,0.994,0.695,0.587,0.725,0.858,0.969,0.927.The research indicated that the variable combination cluster analysis could fully consider the influence between each variable set,and the LSSVM model combined with some best wavelengths and meat texture characteristics was the best method in predicting the water holding capacity and texture profile characteristics of the Tan-mutton.(2)Hyperspectral imaging technology predicted water-holding quality and texture profile characteristics of Tan-mutton during chilled storageHyperspectral imaging technology was used to obtain hyperspectral image and spectral information of chilled mutton samples in seven-days.A prediction model for the water-holding and texture profile characteristics of Tan-mutton during the chilled storage based on NIR hyperspectral imaging technique was established.The best indexes(hardness,adhesiveness and chewiness)were optimized,which could be used to characterize the sensory quality changes of Tan-mutton.Three kinds of pretreatment methods(MC,OSC,OSC+MC)were used to eliminate the spectral data,and the optimal pretreatment combination:MC+OSC pretreatment X and MC pretreatment Y showed the expected results.The VCPA algorithm was used to perform secondary screening on the characteristic spectra filtered by SPA,VCPA and iVISSA methods,which greatly simplified the model operation process.Then established a rapid prediction models of PLSR and LSSVM for hardness,adhesiveness and chewiness of Tan-mutton during chilled storage in those characteristic bands.And the results showed that the VCPA-LSSVM model works well in predicting chewiness(Rp=0.814,RMSEP=352.85),and the iVISSV-VCPA-LSSVM model could predict hardness and cohesiveness(Rp=0.729,RMSEP=3606.807 and Rp=0.856).,RMSEP=938.134)more effective.The research demonstrated that the generalization ability of the model could be improved by the LSSVM algorithm,which effectively explain the dynamic changes of hardness,adhesiveness and chewiness of chilled Tan-mutton during the storage.(3)Hyperspectral imaging technique was used to determine the Tan-mutton chilled storage timeThe near-infrared hyperspectral imaging system in 900-1700nm range was used to collect the spectral information of the mutton samples from the chilled storage.The region of interest was selected,the original spectral information was extracted,the variation of internal quality and spectrum of mutton samples in different chilled storage time were analyzed.Grade the texture changes of the mutton samples during storage by using the spectral change eigenvalue;a rapid PLSDA discriminant model for the chilled storage time was established based on fusion spectrum-image features.The results showed that the accuracy of the calibration set of the PLSDA model of the mutton samples during the chilled storage period of 900-1700nm was 81%,and the accuracy of the prediction set was 73%.The model could be used for discrimination the storage time.