Monitoring The Degradation Of Protein And Fat In Fish During Cold Storage Using Hyperspectral Imaging Technique

Protein and fat are the main organic-nutritious macromolecules in fish,but the degradation of protein and the oxidation of fat during the period of storage would make the loss of fish quality.Traditionally,the chemical methods applied to detect protein and fat are timeconsuming,tedious and sample-destructive;while hyperspectral imaging(HSI)technique can rapidly,non-destructively and synchronously determine the information of protein and fat,and as an advanced techninque that gathering the superiority of spectra and images,it can accelerate the development of food industry.Grass carp and salmon,representing white meat fish and red meat fish respectively,were chosen as the objects for analyzing by hyperspectral imaging technique(400-1000 nm).In the study,the relationship between chemical components and spectra was discussed,and models for predicting protein,fat and fatty acid content in grass carp and salmon were established by fusing multiplex chemometric methods;at the same time,predicted models of TVB-N value and TBA value,revealing the degree of protein degradation and fat oxidation respectively,were built successfully as well.The HSI method,employed to monitor the changes of main organic-nutritious macromolecules,has been established in this study.The main results are as follows.The protein content of fish has been successfully predicted by employing HSI and PLSR.Based on optimal wavelengths selected by RC and texture information selected by GLCM,PLSR models were established and the best performance models of grass carp and salmon have R2 P of 0.811 and 0.825,respectively.The fat and fatty acid content of fish have been successfully determined by HSI and multiplex chemometric methods.For fat content prediction,based on optimal wavelengths selected by SPA and chemical indicators,both PLSR and LS-SVM were used to bulid quantitative analysis models.Among above models,the best prediction model of grass carp was SPA-PLSR(R2P = 0.799),while that of salmon was SPA-LS-SVM(R2P = 0.843).For three unsaturated fatty acid predictions(EPA,DHA and linoleic acid),full spectral wavelengths were optimized by MSC,SNV and 1st derivation,and the simply models fusing selected wavelengths were established by PLSR,MLR and PCR.Among above models,MLR models have the best performance for predicting EPA,DHA and linoleic acid content in both grass carp and salmon,and R2 P of these models are above 0.8,except the model for salmon EPA content prediction has R2 P of 0.786.At the same time,the chemical maps were successfully established.The distribution maps,which reflecting the degree of protein degredation and fat oxidation in fish dynamically during the period of storage,have been successfully established by HSI.For TVB-N value prediction,the best performance model of grass carp and salmon was SPALS-SVM and SPA-BP-ANN,respectively,while RC-MSC-MLR was chosen as best model for both grass carp and salmon in TBA value determination.The most important,the comprehensive models for detecting TVB-N value and TBA value in both white and red meat fish have been successfully built based on spectral information of grass carp and salmon.After comparing,SPA-LS-SVM and RC-MSC-MLR were selected as the best comprehensive models for TVB-N value and TBA value detections,and they have R2 P of 0.907 and 0.842,respectively.At the same time,the chemical maps reflecting single indicator in various objects simultaneously have been plotted.On the basic of above results,it indicated that a simultaneous predicted method for single indicator in various objects by HSI has been successfully established.