Research On The Evaluation Methods Of Chilled Mutton Freshness Based On Multi-information Fusion

Mutton has the unique of delicious taste and rich nutrition,which is favored by consumers.Chilled mutton in non-frozen storage and transportation mode retains the original taste and nutrition of meat to the maximum extent,and has gradually become the mainstream of the meat consumer market.However,the chilled mutton is susceptible to the influence of microorganisms during storage and transportation,resulting in a decrease in freshness fast speed and short shelf life.Freshness is an important standard to measure the economic value and edible of mutton.Accurate detection of freshness can not only protect the rights and interests of consumers and food safety,but also strengthen the supervision of food supervision departments.The traditional methods of meat freshness detection cannot meet the needs of rapid non-destructive detection in Mutton circulation.Due to the complex process of mutton spoilage,a single non-destructive detection method is difficult to get a comprehensive and accurate evaluation.Therefore,it is necessary to study the rapid detection and comprehensive evaluation methods of mutton freshness.In this study,chilled mutton with different freshness was experimentally measured,and the variation rule of freshness and the correlation between freshness indexes during mutton deterioration were analyzed,and the key indexes that could reflect freshness were found out.The gas sensor technology and near infrared spectroscopy technology were used to obtain the microenvironment gas information and spectral information respectively,and the volatile basic nitrogen(TVB-N)content was used as the key index to establish the prediction model of lamb freshness.On this basis,using the way of feature level data fusion processing multisource information,and in three typical methods of artificial intelligence to build based on multi-source data fusion of cold fresh mutton freshness prediction model through the discussion of the multi-source data fusion model,in order to achieve the cold fresh mutton freshness,the purpose of accurate comprehensive evaluation method to explore for the next step of mutton freshness accurate theoretical basis for the development of the nondestructive testing system.The following shows the research content and conclusion:(1)Correlation analysis of fresh degree index of chilled mutton.The mechanism of the corruption and deterioration of cold fresh mutton was studied in depth at different storage time,the physical and chemical indexes,microbial indexes and sensory indexes which affected the freshness of mutton were analyzed.The correlation between the change rules and indexes in the process of mutton corruption was studied.TVB-N was determined as an important index to characterize the freshness of mutton.(2)Freshness evaluation of chilled mutton based on gas sensing information.The multi-sensor gas acquisition device was used to monitor the change of microenvironment gas concentration during the storage of chilled mutton,and the relationship between the concentration of microenvironment gas and the freshness index of chilled mutton was fully explored.With continuous time sequence of mutton oxygen,carbon dioxide and hydrogen sulfide content in the storage environment as the foundation,and TVB-N as freshness evaluation standard,established the mutton freshness prediction model based on BP neural network.Through the above results,the relative error is controlled below 10%,which can monitor the changes of mutton freshness in real time,and basically meets the demand of prediction.(3)Evaluation of chilled mutton freshness based on near infrared spectral information.The spectral information of the chilled mutton at 680nm～2600nm was obtained by NIR spectrometer,and the original data was processed by multiple scattering correction(MSC)method to reduce the spectral error.Then use the successive projections algorithm(SPA)to narrow the gap in existence of information redundancy,and 38 features of selected variables as the BP neural networks and partial least squares regression(PLSR)model input,build TVB-N content prediction model.Through comparing the prediction set as a result,the correlation factor based on BP neural network model and the root mean square error were 0.81 and 3.04mg/100 g,the determination coefficient and root mean square error of PLSR model are 0.78 and 3.15 mg/100 g respectively,indicating that the prediction effect of BP neural network model is obviously better than that of PLSR model.(4)Freshness evaluation of chilled mutton based on multi-source data fusion.The gas data and spectral data obtained by two different technologies are fused by using the feature layer data fusion method.The optimal principal component number obtained by principal component analysis(PCA)was taken as the input value of the prediction model.Then the prediction model of TVB-N content was established based on BP neural network,radial basis function neural network and least squares support vector machine(LSSVM)method,respectively,to realize the comprehensive evaluation of the freshness of cold mutton.By comparing the results showed that two kinds of detection technology of information fusion based model,obtain the single information than one more advantage of the proposed model,three kinds of fusion model to predict the corresponding set of determination coefficient was 0.75,0.85,0.83,respectively,root mean square prediction error was 2.62 mg/100 g,2.39 mg/100 g,2.42 mg/100 g,respectively.The RBF neural network modeling method for the optimal,can more fully and accurately reflect the mutton freshness.