摘要
二维非接触人体测量系统中,围度尺寸需要通过二维至三维的数据转换间接获得。为增强围度预测的精度,应按不同体型类别分别建立预测模型。为避免指标间的信息重复,进行主成分分析,将原有的24项指标归纳为4项综合指标,对人体体型进行客观的综合评价。应用模糊C均值聚类分析进行人体体型分类,分别建立预测模型,以最大隶属度原则对样本进行分类。应用F统计量和聚类有效性函数检验聚类的有效性,并结合实际应用结果确定最佳分类数。
In the 2D non-contacted body measurement, the size of body girths can only be acquired indirectly by data transform from 2D to 3D. In order to increase the precision of the prediction of girth size, it is necessary to develop different prediction models in accordance with body types. To avoid the information overlapping, the principal component analysis is undertaken to sum up the initial 24 independent variables and reduce them to 4 independent variables for objective and comprehensive evaluation of body types. The function of fuzzy clustering analysis is used to make classification of body types whose prediction models are established respectively. The samples are classified on the principle of the maximum membership and the classification cluster validity are verified by F-statistic and the value of clustering validity function. And the optimal number of classification is determined through practical application.
出处
《纺织学报》
EI
CAS
CSCD
北大核心
2007年第2期100-103,共4页
Journal of Textile Research
关键词
主成分分析
体型分类
聚类分析
principal component analysis
classification of body types
clustering analysis
