【摘要】 收集了一年内不同月份不同种类的纯奶牛精补料20个,制备土霉素含量不同的掺假奶牛精补料100个,在全光谱范围内对样品进行近红外透反射光谱扫描,利用CARS法对光谱数据进行前处理,采用偏最小二乘-线性判别分析(PLS-LDA)法来建立判别模型。建立的PLS-LDA模型的交互验证最小错误率为0.0729,模型错分率为0,模型预测错误率为0.0417。说明利用近红外光谱技术建立定性判别模型来检测奶牛饲料中是否掺有土霉素是可行的。更多还原

【Abstract】 In this paper,20 pure dairy concentrate supplements were collected within one year in different months and different types,and 100 oxytetracycline adulteration samples with different contents were prepared.Scan samples with the near-infrared transmission and reflection spectra in the full spectral range,use Competitive Adaptive Reweighted Sampling(CARS) to pretreat spectroscopic data and build the discriminant models with Partial Least Squares-Linear Discriminant Analysis(PLS-LDA).The minimum error rate of cross validation for the established PLS-LDA model was 0.0729,model misclassification rate 0,and the model prediction error rate 0.0417.It is feasible to establish qualitative analysis models to detect whether mixed with oxytetracycline in dairy feed by near infrared spectroscopy.更多还原