Rapid Determination Of The Moisture Content In Dried Scallop Based On Hyperspectral Imaging

Scallops are popular seafood products in the international market owing to its high nutritional value and delicious taste.A large number of scallops need to transport from coastal area to inland.Due to the long-distance transport,it is easily spoilage.Therefore,aquaculture factories often dehydrate scallops to meet consumer's demands,as this prolongs shelf-life and result in high quality products.In this paper,the hyperspectral imaging technique was applied to rapidly determine the moisture content of scallops in different dehydration periods and visualization the distribution of moisture content.Different methods were compared to eliminate the model failure caused by different varieties,providing technique support for digital aquaculture factories.Firstly,mean spectral values of scallops were extracted from hyperspectral images.Combining with the sample set partition methods,such as RS,KS,SPXY and CG,the quantitative models(PLSR and LS-SVM)of moisture content in dried scallop were established and compared.The RS method was suggested for moisture content prediction and visualization map research.Then,the 8 optimal wavelengths of the calibration set divided by RS method,which were selected using the regression coefficient method(RC)and used to establish simplified PLSR and LS-SVM models.After comparing the regression models based on full wavebands and selected wavelengths,a PLSR model based on 8 selected wavelengths(RC-PLSR)was conducted,which produced good performance with Rc = 0.9339,RMSEC = 4.1567%,RP=0.9673,RMSEP =3.5584%,and RPD = 3.7150.Furthermore,based on the optimal RC-PLSR model,the moisture content in each pixel was calculated by inputting the corresponding spectrum into the suggested model and generating a moisture content map.The correlation coefficient between measured and predicted moisture content obtained from the visualization map was 0.9576,implying the potential of visualization maps in describing the spatial distribution of moisture content,along with morphological change.Finally,quantitative and qualitative analysis of the distinction of different scallops,which was aimed to emphasize the necessary of model maintenance.Then studying on the stablility of the optimal model RC-PLSR based on spectrum strandardization.Compared with VSWS-PDS,the results of R = 0.7031 and RMSE = 9.7481%were better than that of VSWS-PDS,indicating that it could effectively eliminate the model instability which was caused by different varieties.