基于近红外光谱技术的面粉水分无损检测模型的建立

Establishment of Nondestructive Testing Model of the Moisture Content in Wheat Flour by Near Infrared Spectroscopy

直接对面粉样品进行近红外光谱扫描,采用105℃恒重法测定面粉中水分含量,在不同的光谱数据预处理方式下运用改进偏最小二乘法(MPLS)建立水分含量定标模型,通过比较模型预测效果以确定最佳预处理方法,随后用PCA、PLS、MPLS三种建模方法在最佳预处理方式下建模通过比较模型预测效果以确定最佳建模方法并用验证集对最优模型进行检验。结果显示,运用MPLS法经标准多元散射校正(Standard MSC)与二阶导数处理后的预测结果最优,其交叉验证标准差(SECV)为0.1874%,预测标准偏差(SEP)为0.381%。这表明运用近红外光谱技术可以实现面粉水分含量的快速检测,进而为食品监管部门在对面粉进行市场监控时提供技术支持。

The wheat flour samples were scanned directly by NIRS. The moisture content model was built up with MPLS at different spectral data preprocessing methods, and the optimal preprocessing method was tested by comparing the models prediction performance. Then the models was established by using the PCA, PLS, and MPLS trader the optimal preprocessing method. The best establishing model method was determined by comparing the three models prediction performance, and the optimal model was tested by the test set. Experimental results indicated that the fore cast results of spectral data was optimal at scattering approach by one rank differential coefficient derivative disposal by using the MPLS, its standard deviation of cross validation （SECV） was 0.1874%. Fore cast standard deviation （SEP） was 0.381%. The NIRS technique can realize fast detection of flour moisture content and provide technical support for food regulator when they are monitoring the flour market.