多尺度边缘检测方法提取近红外光谱信息特征的研究

The Study of Feature Extraction of Near-Infrared Spectroscopy with Multi-Scale Edge Extraction Method

利用小波变换中的多尺度边缘检测方法对甲硝唑药片近红外光谱进行了信息特征的提取。通过偏最小二乘法建立所提取的特征与甲硝唑浓度之间的数学模型,并进行精度分析。结果表明,小波变换多尺度边缘检测能够有效分离并提取光谱信息特征,降低噪声,为一种有效的光谱信息特征提取方法,为解决近红外光谱中光谱叠加严重和信息率低等难题提供了一种思路。

The authors utilized a multi-scale edge extraction method to extract feature information of nearinfrared spectroscopy. Applying partial least-squares to characteristic data obtained, the authors set up a linear mathematical model of the correlation between NIR spectrum and metronidazole contents. The result shows that the multi-scale extraction method can separate and extract feature information, and can be seen as an effective method of spectrum information extraction. The present study offers a new solution to the problems in nearinfrared spectroscopy technology.