近红外光谱技术在南极磷虾粉水分、脂肪和蛋白质含量快速检测中的应用  

Application of near infrared spectroscopy(NIR)technology in the rapid detection of protein,fat and moisture content of Antarctic krill(Euphausia superba)meal

作　　者：苗钧魁[1] 张雅婷 刘小芳[1] 于源 冷凯良[1,2] 杨增光 蒋永毅 MIAO Junkui;ZHANG Yating;LIU Xiaofang;YU Yuan;LENG Kailiang;YANG Zengguang;JIANG Yongyi(Yellow Sea Fisheries Research Institute,Chinese Academy of Fishery Sciences,Key Laboratory of Sustainable Development of Polar Fishery,Ministry of Agriculture and Rural Affairs,Qingdao Engineering Research Center for Polar Fishery Resources Development,Qingdao 266071,China;Pilot National Laboratory for Marine Science and Technology(Qingdao),Qingdao 266200,China;Qingdao Future Testing Technology Co.Ltd.,Qingdao 266111,China)

机构地区：[1]中国水产科学研究院黄海水产研究所农业部海洋渔业可持续发展重点实验室青岛市极地渔业资源开发工程研究中心,山东青岛266071 [2]青岛海洋科学与技术试点国家实验室,山东青岛266200 [3]青岛菲优特检测有限公司,山东青岛266111

出　　处：《食品与发酵工业》2022年第4期243-249,共7页Food and Fermentation Industries

基　　金：国家重点研发计划课题(2018YFC1406804);山东省支持青岛海洋科学与技术试点国家实验室重大科技专项课题(2018SDKJ0304-4)。

摘　　要：利用傅里叶近红外光谱分析技术,以磷虾粉样品的实测值与模型预测值为基础,研究了采用最小二乘法建立磷虾粉原始样品与磷虾粉混合样品中水分、脂肪和蛋白含量近红外定标模型的可行性和准确性。结果表明,磷虾粉近红外图谱最佳预处理方式为:标准正态变换预处理+一阶导数+Norris导数滤波;以磷虾粉混合样品构建的近红外模型较磷虾粉原始样品构建的模型在交互验证均方根误差、外部验证残差均方根(root mean square error of external prediction,RMSEP)(root mean square error of external prediction,RMSEP)和外部验证用样品真实值的标准差与RMSEP的比值(the ratio of the RMSEP to standard deviation of reference data in the prediction,RPD_(EV))等参数有所提升;经预处理后,定标模型的建模相关系数、交互验证相关系数和外部验证相关系数(correlation coefficient in external validation,R_(EV))三类相关系数除脂肪的R_(EV)为0.9058,其余均在0.94以上,RPD均大于2.5,证明磷虾粉近红外定标模型对3个成分均有较好的预测准确性。该研究可为实现南极磷虾粉品质指标的船载近红外快速检测提供参考依据。The Fourier transform near infrared spectroscopy(FT-NIR)analysis technology was used to establish the calibration model based on partial least squares(PLS)for the measurement and predication of Antarctic krill meal samples.The feasibility and accuracy of calibration model for predicting moisture,fat and protein content in atarctic krill meal samples was explored in this study.The result showed that the optimal pretreatment method for the NIR spectra was standard normal variate transform(SNV)+first derivative(FD)+Norris derivative filter(NDF).Compared with the original samples,root mean square error of cross validation(RMSECV),root mean square error of external prediction(RMSEP)and the ratio of the RMSEP to standard deviation of reference data in the prediction(RPD_(EV))of the calibration model of mixed samples were improved.After pretreatment,the correlation coefficient in calibration(R_(EV)),the correlation coefficient in cross validation(R_(EV))and correlation coefficient in external validation(R_(EV))of the calibration model was above 0.94.However,the R_(EV)of the fat was 0.9058 and RPD was above 2.5.Therefore,NIR technology can be used to quickly,conveniently and accurately determine the moisture,fat and protein content of shrimp meal.This study provided reference data for the onboard application of near infrared spectroscopy(NIR)technology for rapid determination of the quality of antarctic krill meal.

关 键 词：磷虾粉 傅里叶近红外光谱 定标模型 快速检测 

分 类 号：O657.33[理学—分析化学] TS254.7[理学—化学]