棉/涤混纺织物近红外简化分析模型构建方法研究  

Research on the Method of Constructing a Simplified Near Infrared Analysis Model for Cotton/Polyester Blended Fabric

作　　者：张国丽[1] 王武[1] ZHANG Guoli;WANG Wu(Tianjin University of Technology Zhonghuan School of Information,Tianjin 300380,China)

机构地区：[1]天津理工大学中环信息学院,天津300380

出　　处：《传感技术学报》2021年第12期1651-1655,共5页Chinese Journal of Sensors and Actuators

基　　金：天津市科技发展战略研究计划项目(18ZLZXZF00480)。

摘　　要：纺织品成分标签是纺织品技术法规严格控制的内容之一。论文采用近红外光谱分析技术对棉/涤混纺织物中棉含量进行分析。标准正态变量变换(SNV)用于消除由颜色、纹理和厚度等差异引起的纺织品光散射和光谱基线漂移;蒙特卡罗无信息变量消除(MCUVE)、竞争性自适应重加权算法(CARS)和迭代保留信息变量(IRIV)三种特征变量优选方法结合偏最小二乘回归(PLSR)用于棉含量近红外简化模型的构建;实验结果表明,经IRIV变量优选方法处理后,建模变量数减少到6,PLSR模型的R;和RMSEP分别为0.909和9.721。该方法为开发纺织品成分便携式近红外分析仪器的光源和探测器选用提供理论参考。Textile composition labeling is one of the contents strictly controlled by textile technical regulations.Near-infrared spectroscopy(NIRS)is used for quantitative analysis of cotton content in cotton/polyester blended fabrics.The standard normal variable(SNV)is used to eliminate light scattering and spectral baseline drift effects caused by differences in color,texture and thickness among textiles.Three characteristic variable optimization methods including Monte Carlo uninformative variable elimination(MCUVE),competitive adaptive reweighted sampling(CARS)and iteratively retaining informative variables(IRIV)combined with partial least square regression(PLSR)are used to construct the near infrared simplified model of cotton contents.The experimental results show that after IRIV variable optimization,the number of modeling variables was reduced to 6,and the R;and RMSEP of PLSR model were 0.909 and 9.721,respectively.This method provides a theoretical reference for the selection of light sources and detectors for the development of portable near-infrared analysis instruments for textile composition.

关 键 词：近红外 简化模型 变量优选 棉/涤混纺织物 成分标签 

分 类 号：TP212.2[自动化与计算机技术—检测技术与自动化装置] TN219[自动化与计算机技术—控制科学与工程]