一种新型光谱多元分析模式识别方法
发布时间：2018-09-11 09:24
【摘要】：SIMCA采用PCA模型参数和F检验构造计算T2i/T2ucl和Si/Q统计量作为样本分类的新属性,并计算待测样本到各类主成分空间的欧式距离作为判别类别的依据,是一种最常用和优秀的光谱分类方法。但是,在Q对T2作图平面上,以欧式距离确定的样本分布范围是一个圆,多数情况下并不一定能符合实际样本分布规律。本文在分析了SIMCA理论缺陷的基础上,提出了一种新方法,即用马氏距离代替欧氏距离作为判别依据来判断样本的类别。并设计了采用红外光谱判别组分比例很接近的掺假食用油样本的实验,以及用近红外光谱判别相近皮毛样本的实验。用调和比5%~8%的食用油红外光谱PCA模型,分别以马氏距离和欧式距离计算出其样本的分布范围,结果表明马氏距离的分类与识别能力更强。新方法和SIMCA对动物皮毛样本的正确识别率分别为87.5%和75%,对比例相近的食用油调和油的正确识别率分别为65%和55%。结果表明新方法对化学组成差异微小的样品分类精度明显优于SIMCA。
[Abstract]:SIMCA uses PCA model parameters and F test to construct T2i/T2ucl and Si/Q statistics as new attributes of sample classification, and calculates the Euclidean distance from samples to various principal component spaces as the basis for classification. It is one of the most common and excellent spectral classification methods. However, the range of sample distribution determined by Euclidean distance is a circle on the map plane of Q pair T2, and in most cases it does not conform to the law of actual sample distribution. On the basis of analyzing the defects of SIMCA's theory, a new method is proposed in this paper, which is to use Markov distance instead of Euclidean distance as the basis to judge the classification of samples. The experiment of using infrared spectrum to distinguish the samples of adulterated edible oil with very close proportion of components and the experiment of using near infrared spectrum to distinguish the samples of similar fur were designed. The PCA model of infrared spectrum of edible oil with a harmonic ratio of 5% and 8% is used to calculate the distribution range of the samples from Markov distance and Euclidean distance respectively. The results show that the classification and recognition ability of Markov distance is stronger. The correct recognition rates of the new method and SIMCA for animal fur samples were 87.5% and 75%, respectively. The correct recognition rates for edible oil blending oil with similar proportion were 65% and 55%, respectively. The results show that the classification accuracy of the new method is better than that of SIMCA. for samples with slight difference in chemical composition.
【作者单位】： 北京化工大学信息科学与技术学院;北京化工大学材料科学与工程学院;碳纤维及功能高分子教育部重点实验室;北京市毛麻丝织品质量监督检验站;内蒙古自治区纤维检验局;
【基金】：国家重大科学仪器设备开发专项(2013YQ220643) 北京市自然科学基金项目(4172044)资助
【分类号】：O657.3;TS227
本文编号：2236288
[Abstract]:SIMCA uses PCA model parameters and F test to construct T2i/T2ucl and Si/Q statistics as new attributes of sample classification, and calculates the Euclidean distance from samples to various principal component spaces as the basis for classification. It is one of the most common and excellent spectral classification methods. However, the range of sample distribution determined by Euclidean distance is a circle on the map plane of Q pair T2, and in most cases it does not conform to the law of actual sample distribution. On the basis of analyzing the defects of SIMCA's theory, a new method is proposed in this paper, which is to use Markov distance instead of Euclidean distance as the basis to judge the classification of samples. The experiment of using infrared spectrum to distinguish the samples of adulterated edible oil with very close proportion of components and the experiment of using near infrared spectrum to distinguish the samples of similar fur were designed. The PCA model of infrared spectrum of edible oil with a harmonic ratio of 5% and 8% is used to calculate the distribution range of the samples from Markov distance and Euclidean distance respectively. The results show that the classification and recognition ability of Markov distance is stronger. The correct recognition rates of the new method and SIMCA for animal fur samples were 87.5% and 75%, respectively. The correct recognition rates for edible oil blending oil with similar proportion were 65% and 55%, respectively. The results show that the classification accuracy of the new method is better than that of SIMCA. for samples with slight difference in chemical composition.
【作者单位】： 北京化工大学信息科学与技术学院;北京化工大学材料科学与工程学院;碳纤维及功能高分子教育部重点实验室;北京市毛麻丝织品质量监督检验站;内蒙古自治区纤维检验局;
【基金】：国家重大科学仪器设备开发专项(2013YQ220643) 北京市自然科学基金项目(4172044)资助
【分类号】：O657.3;TS227
【相似文献】
相关期刊论文 前2条
1 赵杰文;陈全胜;张海东;刘木华;;近红外光谱分析技术在茶叶鉴别中的应用研究[J];光谱学与光谱分析;2006年09期
2 张芙蓉;;马氏距离、Savitzky-Golay平滑求导、SIMPLS结合天麻紫外光谱分析天麻素含量[J];分析测试学报;2012年11期
，本文编号：2236288
本文链接：https://www.wllwen.com/kejilunwen/huaxue/2236288.html
教材专著
