中文摘要：

为了检测异常光谱以及判定异常光谱产生的原因,从相似学原理出发,提出带权相似度度量方法,利用光谱的相似度来判别异常光谱。以中药生产过程中的近红外光谱分析为例,利用带权相似度方法并结合特征谱段的相似性,对异常光谱进行判定。实验结果表明,相比常用的相关系数法、夹角余弦法,带权相似度法灵敏性更高,更能反映关键波长范围内的谱图变化,其判定结果与实际情况重合性较好,从而在近红外在线监测过程中具有很强的实用性。

英文摘要：

In order to detect the spectral outliers and discriminate their causes,this paper proposes a weighted similarity measure method based on the similarity theory,using spectral similarity to discriminate the abnormal spectrums.Taking the Near-Infrared Spectral（NIRS） analysis in the process of Traditional Chinese Medicine production for example,and performing the determination of spectral outliers by combining the proposed method and the similarity of spectrum in feature wavelength range.Experiment shows that the proposed method is more sensitive than other ones（correlation coefficient and vector cosine similarity）,can noticeably reflect spectral changes in the critical wavelength range,and the results of determination coincide with the actual situation very well,so this method has strong practicability in the process of online near-infrared monitoring.