摘要
基于化学计量学方法和新型探测器的快速发展,近红外光谱分析技术被广泛的应用工农业生产、环境保护及安全等领域,为物体成分的快速无损检测提供了基础。本文采用化学计量学算法结合近红外光谱分析实验,深入开展了近红外光谱检测技术的基础研究,建立定性及定量模型来确定检测对象的近红外光谱特征波段,找出以无损方式实时、快速检测的优选方法,从而避免使用复杂的宽光谱感知功能,为发展新型的专用近红外光谱检测设备提供一定的设计基础。对平滑处理、一阶微分和二阶微分等光谱预处理方法进行了系统的比较研究,并采用BP神经网络(BP-ANN)及偏最小二乘(PLS)等算法进行特征变量提取,模型的鉴别准确率达到了98%以上。
For the development of the chemometrics and sensors,the near-infrared spectroscopy analysis method(NIR) is used broadly in the fields of agriculture,industry,environmental protection and security,which provides the basic condition for the rapid and nondestructive analyses for the component of things.In this work,the chemometrics methods combined with NIR analysis method were used to build the quantitative and qualitative models.The pretreatment methods were compared systematically,and the moving back propagation-artificial neural network(BP-ANN) and partial least squares(PLS) were used to select the characteristic wavelengths.The model analysis accuracies are above 98%.
引文
[1]Blanco M.;Villarroya I.Trac-Trend.Anal.Chem.2002,21:240.
[2]褚小立.化学计量学方法与分子光谱分析技术.化学工业出版社.2011,259.