摘要
为使用较少指标对新疆棉花进行分类,提出了一种基于多元统计分析法的棉花分类方法。选择47026包棉花样本的回潮率、马克隆值、上半部平均长度等9个指标进行分析,通过K-Means聚类法确定棉花分类的数量,计算Eta系数并筛选分类指标,运用Fisher判别分析对分类结果进行检验。结果表明:使用马克隆值、反射率、杂质面积率、杂质颗粒数4个指标可将样本分为5类,准确率可达98%;可实现新疆棉花的5级分类,为配棉提供参考。
In order to classify Xinjiang cotton with less indexes,a cotton classification method based on multivariate statistical analysis was proposed.Nine indexes,such as moisture regain,micronaire and upper half mean length of 47026 bales of cotton samples were selected for analysis.The number of cotton classifications was determined by K-Means clustering method,and the Eta coefficient was calculated to screen the classification indexes.Fisher discriminant analysis was used to test the classification results.The results show that the samples can be divided into five categories by four indexes:micronaire,reflectivity,impurity area content and impurity particle number,and the accuracy rate reaches 98%.It can realize the five-level classification of cotton in Xinjiang and provide reference for cotton blending.
作者
梁回香
景建平
阿不都热西提·买买提
张立杰
LIANG Huixiang;JING Jianping;ABUDUREXITI Maimaiti;ZHANG Lijie(Xinjiang University,Urumqi,Xinjiang 830046,China;Xinjiang Autonomous Region Fiber Quality Monitoring Center,Urumqi,Xinjiang 830046,China)
出处
《毛纺科技》
CAS
北大核心
2023年第5期121-126,共6页
Wool Textile Journal
基金
新疆维吾尔自治区科技重大专项(2022A01008-1)。
关键词
新疆棉花
品质指标
聚类分析
相关性
判别模型
Xinjiang cotton
quality index
cluster analysis
relevance
discriminant model
