基于近红外光谱的汽油牌号快速识别  

Fast Recognition of Gasoline Brands Based on Near-infrared Spectroscopy

作　　者：张其可[1] 戴连奎[1] 

机构地区：[1]工业控制技术国家重点实验室

出　　处：《化工自动化及仪表》2005年第4期53-57,共5页Control and Instruments in Chemical Industry

基　　金：国家高技术研究发展计划项目(2003AA412110)

摘　　要：提出基于近红外(NIR)光谱的汽油牌号快速识别算法,主要包括预处理、特征提取和分类建模几部分,比较了各种分类方法的识别能力。实验结果表明:采用主元分析(PCA)提取特征进行模式识别的性能普遍优于直接在光谱波长域的方法,通过选择合适的PCA主元可以获得满意的分类效果,可用于汽油产品的牌号快速识别。The fast recognition algorithms of gasoline brands based on the near-infrared （NIR） spectroscopy is presented. It includes pretreatment, feature extraction and classification modeling. The recognizing capabilities of different kinds of classification methods are compared. Experiment results show that capabilities of recognition methods using principal component analysis （PCA） to extract features are generally better than those used directly in spectrum wavelength domain. The satisfied classification performance can be obtained through selecting proper PCA factors, so it is used in fast recognition of brands of gasoline products.

关 键 词：主元分析 近红外光谱 汽油牌号 模式识别 分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]