Rapid Discrimination Of The Shrimp Freshness Based On Near Infrared Spectroscopy

The quality of aquatic products was directly linked to people’s life and health, and alsoaffected trade of aquatic products, especially the shrimp products was popular for its rich innutrition and delicious. Therefore, it was very essential to assess the freshness and ediblesecurity of aquatic products. As a convenient, fast and high-efficient inspection technology,the Near Infrared Spectroscopy (NIR) had unique features and bright future in theinspection of crops, fruit and vegetables, meat products and other objects. It was veryimportant practical significance that NIR was applied to detect rapidly the quality of aquaticproducts.The freshness of Penaeus vannamei Boone stored at three temperatures (5℃,-18℃and-3℃) was analyzed in this paper by near infrared spectroscopy analyzer (DA7200). Studiedwas the near infrared spectral curve of complete shrimp and minced shrimp during storage,discussed was the near infrared spectral curve variation of complete shrimp and mincedshrimp for different storage temperature and time. Based on sensory evaluation, volatilebase nitrogen (TVB-N) and total bacterial count((TBC),the mathematical model wasestablished to determinate the Penaeus vannamei Boone’s content of TVB-N and TBC bynear infrared spectroscopy, and the partial least square regression (PLSR) was used to buildTVB-N and TBC model to distinguish shrimp freshness.The main results were as follows:(1) Analysis of sensory score, physical and chemical indexes and microbial indexes ofPenaeus vannamei Boone in different storage temperature. The results showed that sensoryscore, the value of TVB-N and TBC increased with the storage time, and the higher thestorage temperature, the faster the increase.(2) The near infrared spectral curve of complete shrimp and minced shrimp wascollected and analyzed to discriminate shrimp freshness. The results showed that thespectral curve showed certain regularity for different storage temperature and time. thefigures of the two dimensions were drawn with the ratio（A980nm/A1450nm）of peak parameterin980nm and1450nm as the X axis，the ratio(A1200nm/A1350nm) of peak parameter in1200nm and1350nm as the Y axis, shrimp freshness could be correctly discriminated, andthe results of minced shrimp was better than the complete shrimp, the results of shrimpstored at5℃was the best, and the results of shrimp at-3℃was better than that at-18℃.(3) The value of TVB-N and TBC was quantitative analyzed. The whole samples ofcomplete shrimp and minced shrimp were scanned by a spectroradiometer, the spectra wasobtained range from950to1650nm. The raw spectra were pretreated by first derivative,Savisky-Golay (SG) smoothing and Standard Normal Variate (SNV), the range from950to1450nm was chose using the Unscrambler10.3. The mathematical model was established to determinate the shrimp’s content of TVB-N and TBC by PLSR, and was validated. Theresults showed that, the evaluation index in calibration set of model were the correction ofdecision Rc2, the correlation coefficients rc, the root mean square error correction RMSEC,standard error of calibration SEC, the correction of decision of Cross-Validation Rc2, thecorrelation coefficients of cross-validation rv, root mean standard error of cross-validationRMSECV, standard error of cross-validation SECV, and the evaluation index in predictionset of model were the correlation coefficients, the root mean square errors of predictionRMSEP, standard error prediction SEP, the deviation of measured and predicted value.Results indicated that these models were better to predict TVB-N and TBC in the Penaeusvannamei Boone，and could distinguish shrimp freshness and the results were consistentwith the discrimination results of sensory score, the value of TVB-N and TBC. Theevaluation index showed that the minced shrimp models of TVB-N and TBC were betterthan complete shrimp models and the whole shrimp samples models. The correlationcoefficients of TVB-N and TBC in prediction set of the best models were0.961and0.943respectively, RMSEP were2.179and0.603respectively. Results showed that the modelswere better and could be used for quantitative analysis.