基于高光谱的云岭牛雪花牛肉氨基酸含量预测  

Hyperspectral prediction of amino acid content in Yunling marbled beef

作　　者：罗爽 杨林楠[1,3,4] 张丽莲 彭琳[1,3,4] 李佩杉 郜鲁涛[1,2,3,4] LUO Shuang;YANG Lin-nan;ZHANG Li-lian;PENG Lin;LI Pei-shan;GAO Lu-tao(College of Big Data,Yunnan Agricultural University,Kunming 650201,China;College of Food Science and Technology,Yunnan Agricultural University,Kunming 650201,China;Yunnan Engineering Technology Research Center of Agricultural Big Data,Kunming 650201,China;Yunnan Engineering Research Center for Big Data Intelligent Information Processing of Green Agricultural Products,Kunming 650201,China)

机构地区：[1]云南农业大学大数据学院,昆明650201 [2]云南农业大学食品科学技术学院,昆明650201 [3]云南省农业大数据工程技术研究中心,昆明650201 [4]云南省绿色农产品大数据智能信息处理工程研究中心,昆明650201

出　　处：《湖北农业科学》2024年第7期120-128,共9页Hubei Agricultural Sciences

基　　金：云南高原优质肉牛产业智能化管理研究与示范项目(202102AE090009);云南省基础研究专项-面上项目(202101AT070248)。

摘　　要：为建立一种基于高光谱成像技术结合机器学习的雪花牛肉氨基酸含量无损、快速测定的方法,采集云岭牛5个等级100组的雪花牛肉分别在400~1 000 nm和900~2 500 nm波段高光谱数据,使用JJG1064-2011标准氨基酸分析仪测定样本中17种氨基酸含量;采用一阶差分(1st Derivative,D1)进行高光谱数据预处理,使用连续投影算法(Successive projection algorithm,SPA)提取特征波段。采用决策树(Decision trees)、支持向量机(Support vector machine,SVM)、岭回归(Ridge regression)、偏最小二乘回归(Partial least squares regression,PLSR)以及卷积神经网络(Convolutional neural network,CNN)5种方法预测氨基酸含量。结果表明,结合D1预处理、SPA特征提取建立CNN模型在预测氨基酸含量方面表现最佳,其均方误差(Mean squared error,MSE)为0.010 3,平均绝对误差(Mean absolute error,MAE)为0.082 2,决定系数(Coefficient of determination,R2)为0.898 5。A method for non-destructive and rapid determination of the amino acid content of Yunling marbled beef based on hyper-spectral imaging technology combined with machine learning was introduced.Hyperspectral data were collected in the 400~1000 nm and 900~2500 nm bands for 100 groups of marbled beef from five grades of Yunling cattle.The JJG1064-2011 standard amino acid analyzer was used to measure the content of 17 amino acids in the sample.The first-order difference(1st Derivative,D1)was used for hyperspectral data preprocessing,and the Successive projection algorithm(SPA)was used for feature band extraction.Five methods including Decision trees(Decision trees),Support vector machine(SVM),Ridge regression(Ridge regression),Partial least squares regression(PLSR)and Convolutional neural network(CNN)were used for predicting amino acid content.Experimental re-sults showed that the CNN model combined with D1 preprocessing and SPA feature extraction performed best in predicting amino acid content,with mean squared error(MSE)of 0.0103,mean absolute error(MAE)of 0.0822,and the coefficient of determination(R2)of 0.8985.

关 键 词：高光谱成像技术 云岭牛雪花牛肉 氨基酸 预测模型 

分 类 号：O157.3[理学—数学]