RBF模糊神经网络用于NIR鉴别羊绒和羊毛的可行性研究

为了实现羊绒、羊毛纤维的快速、无损检测,建立了羊绒、羊毛近红外光谱数据库,包括228组各地羊绒、羊毛数据,并应用于羊绒、羊毛的定性检测上。首先介绍了羊绒、羊毛近红外光谱检测的数据库建立过程;然后,在对羊绒、羊毛原始近红外光谱进行预处理的基础上,对数据进行主成分分析,选出12种主成分,并结合改进的RBF模糊神经网络,建立羊绒、羊毛检测模型。通过与主成分分析-马氏距离建模方法的对比分析实验表明,建立近红外光谱数据库,并结合主成分分析和改进的RBF模糊神经网络的方法是一种有效的无损检测羊绒、羊毛的方法,可快速建立高精度的羊绒、羊毛纤维检测模型。

Feasibility Study of RBF Fuzzy Neural Network in Cashmere and Wool Identification Based on Near Infrared Spectroscopy

In order to realize the fast and nondestructive detection,the cashmere and wool near infrared spectroscopy database is created which includes the data of 228 groups of cashmere and wool from various districts,and it is applied to the qualitative detection of cashmere and wool.First the process of database creation in the cashmere and wool detection based on near infrared spectroscopy is introduced.Then on the base of the near-infrared spectroscopy original data preprocessing of cashmere and wool,the principal components of the data are analyzed,and 12 kinds of principal components are chosen.The detection model of cashmere and wool with radial basis function(RBF) fuzzy neural network is build.The comparative analysis experiments with PCA-MD modeling demonstrate that the method combining near infrared spectroscopy database,principal components analysis(PCA) and improved RBF fuzzy neural network is an effective and nondestructive detection method for cashmere and wool,and it can rapidly build high-accuracy detection models of cashmere and wool fiber.