不同部位驼肉营养成分检测与近红外光谱快速预测模型建立

The detection of nutrient components in different parts of camel meat and the establishment of near infrared rapid prediction model

该实验选取3~5岁龄阿拉善双峰驼胴体的辣椒条、骆驼霖、大黄瓜条、腹肉、腱子肉、里脊、米龙、上脑、臀肉、外脊、胸肉、小黄瓜条、眼肉等13个部位肉样,测定驼肉水分、蛋白质和脂肪含量,并利用近红外光谱仪在4000~10000 cm^(-1)波数条件下进行光谱扫描。首先选用滤波平滑(savitzky-golay,S-G)、导数、标准正态变量转换法(standard normal variate,SNV)、多元散射校正(multiplication scatter correction,MSC)方法进行光谱预处理比较,其次采用偏最小二乘判别分析法(partial least square-discrimination analysis,PLS-DA)建立水分、蛋白质、脂肪含量模型。结果表明,不同部位驼肉样的水分含量模型的最佳预处理方法为MSC,模型的决定系数R_(c)^(2)、校正集均方根误差(root mean square error of calibration,RMSEC)、R_(p)^(2)、验证集均方根误差(root mean square error of prediction,RMSEP)分别是0.7459、0.0085、0.7741、0.0134;蛋白质含量模型的最佳预处理方法为SNV,模型的决定系数R_(c)^(2)、RMSEC、R_(p)^(2)、RMSEP分别是0.6602、0.2879、0.6725、0.2760;脂肪含量模型的最佳预处理方法为S-G,模型的决定系数R_(c)^(2)、RMSEC、R_(p)^(2)、RMSEP分别是0.9885、0.0863、0.9963、0.0567。研究表明利用近红外光谱技术对不同部位驼肉样脂肪含量的预测最好,其次为水分含量的预测,蛋白质含量预测的结果最不佳。

In this experiment,we selected pepper strips,camel lin,large cucumber strips,belly meat,tendon meat,tenderloin,milong,upper brain,rump,outer spine,breast meat,and small cucumber from the carcass of Alashan Bactrian camel aged 3-5 years old.Thirteen meat samples such as strips and eye meats were scanned by a near-infrared spectrometer under the wavenumber condition of 4000-10000 cm^(-1),and the moisture,protein and fat content of the camel meat samples were also been measured.First,savitzky-golay(S-G),derivative(Der),standard normal variate(SNV),and multiplication scatter correction(MSC)methods were selected for com-paration of the spectral preprocessing.Then,partial least square-discrimination analysis(PLS-DA)was used to establish water,protein,and fat content models.The results showed that the best pretreatment method for the moisture content model of different parts of camel meat samples was MSC.The determination coefficients R_(c)^(2),root mean square error of calibration(RMSEC),R_(p)^(2),and root mean square error of prediction(RMSEP)of the model were 0.7459,0.0085,0.7741,0.0134,respectively.The best pretreatment for the protein content model was SNV,and the determination coefficients R_(c)^(2),RMSEC,R_(p)^(2),and RMSEP of the model were 0.6602,0.2879,0.6725,and 0.2760,respectively.The best preprocessing method of the fat content model was S-G,and the determination coefficients of the model R_(c)^(2),RMSEC R_(p)^(2),and RMSEP were respectively 0.9885,0.0863,0.9963,0.0567.The results showed that near-infra-red spectroscopy technology was the best way to predict the fat content of different parts of camel meat,followed by the prediction of mois-ture content,and the prediction of protein content was the worst.