Research On Detection Technology Of Fish Freshness And Morphological Parameters For Production Line

At present,due to the limitation of processing equipment,fish are usually directly frozen to maintain freshness after fishing.However,the preservation temperature cannot be kept constant during transit shipment,so there are multiple freezing-thawing phenomena that seriously affect the freshness.Moreover,to extend the shelf life of fish,evisceration must be carried out,due to the very perishable viscera of fish.Unfortunately,the automatic fish processing system is imperfect,and the processing mode is single.It is impossible to predict its internal morphological parameters according to the type and size of the fish body.Therefore,the paper focuses on the rapid non-destructive testing of fish freshness and morphological parameters on the production line.（1）In order to realize the fast non-destructive detection of the fish body shape parameters oriented to the processing line,the MRI technology is used to preprocess the image to establish the fish body DICOM format data set.Manual segmentation,interactive segmentation and regional growth segmentation are used to segment and reconstruct the fish body,and the external morphological parameter data of the fish body are obtained.The 7 external morphological parameters of fish body extracted as independent variables.Multiple linear regression（MLR）and least squares regression（PLSR）were used to establish a fish body quality prediction model based on the external morphological parameters.Using the measured actual visceral volume(1(1₁and standard ellipsoidal visceral calculated volume(1(1₂ for regression analysis,a fish visceral volume prediction model is established and corrected.Combined with multiple internal linear regression models of visceral parameters（maximum length L,maximum width W,maximum thickness T）established using 7 external morphological parameters of the fish body respectively,the internal morphological parameter prediction based on the external morphological parameters of the fish body was achieved The coefficient value²₍(8（8） is 0.852 and the value²is 0.827,which has a good prediction accuracy.（2）To realize the freshness detection of frozen and thawed fish,the sensory evaluation method and TPA texture analysis method of conventional freshness detection technology were used to analyze the freshness of the fish body after repeated freezing and thawing 0-5 times.The 8 sensory evaluation indicators were employed to evaluate the freshness of the fish body.Three types of freshness and spoilage,combined with total sensory scores and freezing and thawing times,established sensory freshness evaluation standards.The texture parameters of the fish body were tested by repeated freezing and thawing from 0-5 times.The change trend of texture characteristics is consistent with the result of sensory evaluation,and has a significant correlation.Principal component analysis was used to extract the main components of the texture characteristics,and the PLSR model was used to establish a sensory total score prediction model based on the texture characteristics of frozen and thawed fish.Combined with the sensory freshness evaluation standards,objective evaluation of fish freshness can be achieved.（3）To achieve fast and non-destructive detection of freshness of frozen and thawed fish,hyperspectral detection technology is used to construct a qualitative detection model for repeated freshness of frozen and thawed fish.The corrected spectral image is used to collect the original spectrum of the fish gill ROI region.Then,spectral preprocessing is implemented including the SG smoothing,multivariate scattering correction（MSC）and variable standardization（SNV）.Finally,the continuous projection algorithm（SPA）and competitive adaptive re-weighting algorithm（CARS）are used to filter the characteristic wavelengths to reduce the dimensionality of the spectral data.SPA selected 5 characteristic wavelengths and CARS selected 18 characteristic wavelengths.Aiming at full-spectrum information and characteristic wavelength information,soft independent pattern classification（SIMCA）and partial least squares discriminant analysis（PLS-DA）are used to establish the optimal qualitative identification model of freshness of frozen and thawed fish.