Quality Evaluation Of Eriocheir Sinensis Based On Spectrum And Imaging Technology

Eriocheir sinensis is a famous aquatic product in China.It is favored by consumers because of its rich nutrition and delicious meat.In recent years,its yield and output value are huge and increasing continuously.At present,artificial monitoring and grading are used in the aquiculture and production of Eriocheir sinensis.However,the artificial method is inefficiency and subjectively,which result in low yield in the culture and slow marketing speed.Besides,it is difficult to guarantee the quality and freshness of Eriocheir sinensis.Therefore,it is essential to study on rapid non-destructive detection method and system for monitoring and quality identification of Eriocheir sinensis.In this study,spectrum and imaging technology was used to identify the quality of Eriocheir sinensis.Firstly,this study aims at the application of visible-near infrared（421⁹63 nm）hyperspectral combined with chemometrics method,for rapid nondestructive prediction method of crab shell hardness,chitin and protein during post-shelling hardening.Secondly,studying on a rapid non-destructive prediction method of Eriocheir sinensis grade based on spectral image fusion information.Finally,the Eriocheir sinensis online grading system based on machine vision technology was established to achieve the online grading of sex,quality and weight.The main research contents are as follows:（1）Detection hardness,chitin and protein of crab shell during hardening based on hyperspectral technologyTaking post-shelling hardening Eriocheir sinensis as the research object,the hardness,chitin content and protein content of the crab shell on back were measured,and the visible-near infrared（421-963 nm）spectra of hyperspectral images.After comparing various spectral pretreatment methods(1^stDER,2^ndDER,MSC,SNV,SG),the optimal pretreatment method was selected.Then,siPLS,GAPLS and si-GAPLS were used to screen characteristic wavelengths.Quantitative prediction models of PLS and SVM were established based on full spectrum and characteristic wavelength respectively to achieve fast non-destructive prediction of crab shell hardness,chitin content and protein content.The results showed that hardness index is significant negatively correlated with chitin content and protein content during post-shelling hardening,and GAPLS,si-GASVM and GASVM have the best predictive effects on hardness,chitin content and protein content respectively.（2）Studying on Eriocheir sinensis grade based on spectral image fusion information of hyperspectral technologyTaking Eriocheir sinensis of different quality grades as the research object,the visible-near infrared（421-963 nm）spectra of hyperspectral images and image information of hyperspectral principal component images of the crab shell on back were measured.MSC method was used to pretreat the spectrum,then GA and ACO were used to screen the characteristic wavelength.The classification models of RF,LDA,KNN,BP-ANN and LS-SVM were established based on the characteristic wavelength,image information and fused characteristic wavelength and image information,respectively,to achieve the quality classification of Eriocheir sinensis.The results showed that the recognition effect of RF model based on fusion information is the best,and the recognition rate of prediction set is98.89%.The classification model based on image information and the selected characteristic wavelengths of 486 nm,745 nm and 908 nm can provide theoretical basis for the construction of grading system.（3）On-line classification system of Eriocheir sinensis based on visual features and weightThe ventral images of Eriocheir sinensis and back images of Eriocheir sinensis at 486nm,745 nm and 908 nm were collected respectively.Sex information recognition of Eriocheir sinensis using template matching algorithm for ventral image.The image information on the back under three wavelengths is extracted,and the crab shell area extracted combine with weight to obtain the plumpness.BP-ANN classification model was established based on image information and plumpness.The results showed that the template matching algorithm can effectively identify sex information of Eriocheir sinensis,and the recognition rate is 98.75%.BP-ANN classification model can effectively identify the quality grades of male and female crabs,with recognition rates of 94.38%and 95.63%,respectively.This system achieves online grading of sex,quality and weight of Eriocheir sinensis.In this study,the rapid non-destructive prediction of the hardness of crab shell during the hardening process and the grade of Eriocheir sinensis was realized,and an on-line grading system of Eriocheir sinensis was developed,which is of great significance for improving the level of automatic grading of aquatic products in China.