基于空间向量模长的青年女性腰腹臀部形态分类与判别  

Classification and discrimination of waist-abdomen-hip morphology of young women based on space vector length

作　　者：吴金颖 李炘 丁笑君[1,2,3] 邱文池 邹奉元 WU Jinying;LI Xin;DING Xiaojun;QIU Wenchi;ZOU Fengyuan(School of Fashion Design&Engineering,Zhejiang Sci-Tech University,Hangzhou,Zhejiang 311199,China;Clothing Engineering Research Center of Zhejiang Province,Hangzhou,Zhejiang 311199,China;Key Laboratory of Silk Culture Heritage and Products Design Digital Technology,Ministry of Culture and Tourism,Hangzhou,Zhejiang 311199,China)

机构地区：[1]浙江理工大学服装学院,浙江杭州311199 [2]浙江省服装工程技术研究中心,浙江杭州311199 [3]丝绸文化传承与产品设计数字化技术文化和旅游部重点实验室,浙江杭州311199

出　　处：《纺织学报》2024年第4期180-187,共8页Journal of Textile Research

基　　金：文化和旅游部重点实验室开放基金项目(2020WLB09);国家级大学生创新创业训练计划项目(202210338032)。

摘　　要：为实现女体三维曲面的分类与判别,提高裤装的合体性,提出一种基于空间向量模长表征腰腹臀部形态并进行体型分类的方法。获取323名18~25岁青年女性的三维点云数据,提取腰腹臀部10层横截面曲线,以臀围质心为原点重建点云坐标系,通过欧式距离计算130个特征点的空间向量模长,构建表征人体曲面形态的模长矩阵。引入拉普拉斯特征映射降维获取18个本征维度,采用K-means聚类,运用随机森林算法建立青年女性腰腹臀部形态的判别模型。结果表明,青年女性腰腹臀部可分为大众型、扁平型、腹凸型、臀凸型4类,分别占样本总数的58.82%、27.86%、8.36%和4.95%。分析获得4类体型对应的裤装样板差异,有效提高了裤装合体性,青年女性腰腹臀部形态判别准确率达96.92%。Objective The current research on body shape uses classification indexes for body shape such as circumference,width,thickness,ratio and angle,which are not able to fully reflect the curved shape of body.In order to establish classification and discrimination of three-dimensional surfaces of female body and improve the fit of pants,a method of body shape classification based on space vector length characterization of waist-abdomen-hip morphology is proposed.Method The 3-D point cloud data of 323 young women aged 18-25 years were collected by TC 23D scanner,and the 10-layer cross-sectional curves of the waist-abdomen-hip were extracted.The center of mass of hip circumference was used as the origin to reconstruct the point cloud coordinate system.The space vector length of 130 feature points was calculated by Euclidean distance to construct the modulus length matrix to characterize the morphology of body surface.The eigen dimension was determined by using maximum likelihood estimation,and Laplace feature mapping was introduced.The body type segmentation based on space vector length was achieved by K-means clustering.The discriminative model of waist-abdomen-hip morphology of young women was established by random forest algorithm.The garment samples(SVP)corresponding to four types under size 160/66A were obtained and compared with the basic women's suit pants benchmark sample(BTP)to evaluate their fit.Results Based on the 3-D point cloud data,10 feature cross-sections reflecting the morphology of the waist-abdomen-hip surfaces were extracted,and 130 feature points were extracted.A modal length matrix of 323 samples,each with 130 space vector lengths was constructed.The maximum likelihood estimation was used to determine the eigen dimension as 18.The modal length matrix 32318 characterizing the body surface morphology was obtained by dimensionality reduction.The elbow method was used to determine the number of clusters as 4.Through cluster analysis,the waist-abdomen-hip of young women were subdivided into four types b

关 键 词：青年女性 腰腹臀部形态 空间向量模长 体型分类 合体性 服装 

分 类 号：TS941.17[轻工技术与工程—服装设计与工程]