近红外光谱无损检测技术中数据的分析方法概述  

Overview of Data Analysis Methods in Near-Infrared Spectroscopy Nondestructive Testing

作　　者：刘军[1,2] 吴梦婷[1,2] 谭正林[3] 李威 

机构地区：[1]智能机器人湖北省重点实验室(武汉工程大学),湖北武汉430205 [2]武汉工程大学计算机科学与工程学院,湖北武汉430205 [3]湖北经济学院烹饪与营养学系,湖北武汉430205

出　　处：《武汉工程大学学报》2017年第5期496-502,共7页Journal of Wuhan Institute of Technology

基　　金：湖北省食品药品监督管理局项目(201610+13);湖北省智能机器人重点实验室开放基金(HBIR 201608);武汉工程大学研究生创新基金(CX2016063)

摘　　要：近红外光谱无损检测技术可用于品种鉴别与农产品的定性或者是定量的分析工作.本文介绍了近红外光谱的基本原理及各类近红外光谱分析方法.近红外光谱无损检测技术中数据分析方法是通过光谱定量分析找到光谱以及对应浓度的内在关系,建立相应的数学模型.这些方法主要有偏最小二乘回归、主成分分析法、BP神经网络算法、支持向量机、K最近邻分类算法和线性判别分析法等.通过这些分析模型的对比,研究表明:支持向量机将是近红外光谱数据分析方法未来一个重要的研究方向.Near-infrared spectroscopy nondestructive testing technology can be used for variety identification and the qualitative or quantitative analysis of agricultural products. The basic principle of near-infrared spectroscopy and the methods of near-infrared spectrum analysis were introduced. The data analysis methods in near-infrared nondestructive testing technology aim at finding the relationship between the spectrum and the corresponding concentration through the quantitative analysis of the spectrum, and establishing the corresponding mathematical model, which mainly include partial least squares regression, principal component analysis, back propagation artificial neural network, support vector machine （SVM） , K-Nearest neighbor classification algorithm and linear discriminant analysis. The comparison result of these analytical models show that SVM method may be a future research direction in near infrared spectrum data analysis.

关 键 词：近红外光谱 无损检测 数据分析方法 

分 类 号：R857.3[医药卫生—航空、航天与航海医学]