Study On Chemometrics In Quality Detection And Analysis Of Aquatic Products

According to the Chemometrics, which can choose the optimal measurement method,the most efficient way to obtain useful feature data of system, and its significant featureson extraction of the material qualitative, quantitative, morphology, structure and otherinformation through analyzing measurement data, we tried to find a new quality detectionmethod of aquatic product under useing it to detect and analyse food nutrition and qualitycombining with analytical instruments and computer numerical method.The main research content and results are following:1Application of Chemometrics in the detection of amino acid hydrolysate in threeshrimpThe content of amino acids especially the essential amino acids is an important indexof measuring the nutrition value in protein foods. In our study, we use ELM-ultravioletspectroscopy, BP-ultraviolet spectroscopy and RBF-ultraviolet spectroscopy to foundmodels to get the content of three kinds of amino acids (phenylalanine, tyrosine andhistidine) under amino acid analyzer data of the muscle protein hydrolysate of shrimp body,and compared the superiority of the prediction performance of the three. Results of meansquare error of ELM, BP and RBF prediction are3.9667e-007,7.0938and5.2379e-004.The average value of R2, the decisive coefficient of ELM, BP, RBF are9.42069,8.76012and8.80471. The experimental results showed that BP and RBF have larger errors, butELM has faster learning speed and higher predictive accuracy and can provide referencefor the content detection of amino acids in food and new ideas for food quality analysis.2Study on Chemometrics and computer image processing technology in the rapiddetection of fish freshnessTo explore a new method for rapid detection of crucian freshness; sensoryidentification methods of fish freshness show that changes of fisheye clarity, gloss andcolor have a very significant relationship with their freshness. During the experiment, wetake eyes images of crucian via digital camera every six hour and obtain their signalstrength with computer image processing technology. Analysis relation between changes ofsignal strength and the values(pH, electrical conductivity, TVBN) by modeling of PLS andSVR. We found that complex correlation coefficient R value of signal strength and thevalues(pH, electrical conductivity, TVBN) measured by with the principal component analysis of PLS method is0.9737. It illustrates the signal strength can be an indicator offreshness detection of crucian carp. Obtained values and predicted values of the signalstrength of PLS model all are distributed on the fitting line, correlation coefficient R is09967. The mean square errors of the training set and testing of SVR model are0.00025and0.0012, R2are0.999and0.995. In conclusion, both methods achieve the fast and exactfreshness detection of crucian. The two methods can be usesd for rapid and accuratedetection of crucian freshness.