采用近红外光谱技术的燕麦脂肪含量检测

Determination of fat content of oat by near infrared spectrum

以93份燕麦样品为研究对象,对其近红外光谱数据进行预处理后通过主成分分析法提取光谱特征,采用人工神经网络技术建立燕麦中脂肪含量的合理检测模型。结果表明:反向多元散射处理(IMSC)、数学处理选择2441(即对光谱进行导数间隔点为4的二阶导数处理,一次平滑处理间隔点为4,不进行二次平滑处理)为最佳预处理方法;通过主成分分析法提取2个主成分作为原始信息的特征变量,建立的人工网络模型结构为2-17-1,该模型对验证集的测定值与预测值的相关系数为0.962 3,均方根误差为1.607 2,模型的预测准确性较好。该方法简便、快速,为燕麦脂肪的定量测定提供了一种新方法。

The near infrared spectrum（NIRS）technology is used to detect the fat content of oat.With 93 oat collected from China as samples,the model of the artificial neural network for determining fat content in oat is established after preprocessing the near infrared spectrum data and extracting the spectral characteristics by principle component analysis（PCA）.It is shown that the preprocessing of spectrum scattering is the inverse multiple scatter correction（IMSC）and mathematics processing is 2441（2is the second derivative processing;4is the interval point of the second derivative;4is the first smoothing interval point and 1is no secondary smoothing）.The structure of the artificial neural network mode is 2-17-1,which is established after extracting2 principle component as the characteristic variables of the original information.The correlation coefficient of the true value and the prediction value is 0.962 3,and the root mean square deviation is 1.607 2.The model has better predictive accuracy and can be used to detect fat content in oat rapidly.