样品状态与装样条件对近红外光谱测定的影响研究

近红外谱区的信噪比较低和测定时的复杂背景，导致近红外光谱易受样品状态和装样条件的影响。本文以2组样品集，考察同一样品集在三种不同的制样状态下的漫反射光谱所建的模型；以及预测时出现装样条件变化时对预测结果的影响及相应的预处理方法研究。试验表明：样品的均匀度越好，则所建近红外模型越优；针对在线分析可能出现的装样疏密度变化，矢量归一化可有效地降低装样稀疏的预测误差。

Study on the Influence of Sample''''s State and Loading Condition to Near Infrared Spectroscopy Analysis

Near infrared spectrum (especially the diffuse reflection spectrum of powder or seed) is easily affected by sample's state and loading condition because of the lower Signal Noise Ratio in the spectrum region and the complex back- ground while detecting. The models built by the same sample set in three statesare are compared. And aim at the uncontrol- lable loading condition, Some pretreatment algorithms to eliminate the influenceare are discussed. The experimental result indicates that the near infrared spectrum model has been influenced by the sample's size uniformity and Normalize can effectively eliminate the influence caused by sparsely loading.