两种近红外光谱分辨率预测牛肉营养成分的比较研究

Comparative study on the prediction of beef nutrients by near infrared spectroscopy under two resolutions

应用近红外光谱技术在不同光谱分辨率下分析了同一批牛肉样本的蛋白质、脂肪和水分含量。样品取自16头西门塔尔杂交牛的14个部位,宰后成熟48h,绞成肉糜状后分别于不同分辨率1.6和10.0nm条件下进行近红外光谱扫描和化学成分测定。应用The Unscrambler建模软件,采用偏最小二乘回归技术(PLSR),通过交互验证程序建立近红外数学模型,得到不同分辨率1.6和10.0nm条件下蛋白质的校正集相关系数R分别为0.94和0.93,交互验证标准差(RMSECV)分别为0.49和0.54;脂肪R分别为0.93和0.92,RMSECV分别为0.64和0.76;水分R分别为0.87和0.81,RMSECV分别为1.18和1.26。研究结果表明,高光谱分辨率下的蛋白质、脂肪和水分模型精度要略优于低光谱分辨率所建模型。

The protein,fat and moisture of the beef samples under two resolutions were analyzed using near infrared spectroscopy.The samples were obtained from 14 parts of 16 Simmental crossbred cattle. After 48h postmortem aging,these samples would be homogenized and scanned.Immediately after scanning under 1.6 and 10.0nm by near infrared spectroscopy（ NIR） ,the samples were analyzed for protein,fat and moisture.The models were set up by partial least squares regression（ PLSR） using the Unscrambler software.The results of nutrient contents tested by cross-validation under two resolutions of 1.6 and 10.0nm showed R of 0.94 and 0.93,RMSECV of 0.49 and 0.54 （ protein） ; R of 0.93 and 0.92,RMSECV of 0.64 and 0.76 （ fat） ; R of 0.87 and 0.81,RMSECV of 1.18 and 1.26 （ moisture） ,respectively.The above research results demonstrated that for the models of protein,fat and moisture, the higher resolution provide slightly better results than the lower resolution.