基于二维图像的青年女性颈肩部形态自动识别  

Automatic identification of young women′s neck-shoulder shapes based on images

作　　者：王婷 顾冰菲 WANG Ting;GU Bingfei(School of Fashion Design & Engineering, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China;Clothing Engineering Research Center of Zhejiang Province, Hangzhou, Zhejiang 310018, China;Key Laboratory of Silk Culture Inheriting and Products Design Digital Technology, Ministry of Culture and Tourism, Hangzhou, Zhejiang 310018, China)

机构地区：[1]浙江理工大学服装学院,浙江杭州310018 [2]浙江省服装工程技术研究中心,浙江杭州310018 [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310018

出　　处：《纺织学报》2020年第12期111-117,共7页Journal of Textile Research

基　　金：国家自然科学基金项目(61702461,61702460);中国纺织工业联合会科技指导性项目(2018079);中国纺织工业联合会应用基础研究项目(J202007);浙江理工大学科研业务费专项资金资助项目(2020Q051);浙江理工大学服装服饰文化创新团队项目(11310031282006)。

摘　　要：为实现青年女性颈肩部形态的自动识别,首先基于202名在校青年女性的三维点云数据,测量了15个颈肩部形态相关参数,通过分析确定出离散程度较大的形态参数,包括肩斜角、背入角、肩矢额径比和腋下矢额径比;然后结合这4个重要体型参数,对青年女性颈肩部形态进行细分并建立各类体型的分类规则;最后基于青年女性正面与侧面二维照片,通过提取人体轮廓和识别特征点获得颈肩部体型分类所需参数,根据体型分类规则实现颈肩部形态的自动识别。结果表明:青年女性颈肩部形态可分为4类,即圆宽肩体、扁窄肩体、圆落肩体、驼背扁肩体,分别占样本总数的25.53%、23.94%、25.59%和23.94%;通过对40名测试样本进行基于正、侧面二维照片的颈肩部形态自动识别验证,准确率达到90%,说明基于本文方法构建的颈肩部体型自动识别系统是有效的。In order to facilitate the automatic identification of young women′s neck-shoulder shapes,15 neck-shoulder shape parameters of 202 young women were measured in the form of three-dimensional point cloud data,and the parameters with a large degree of dispersion were determined through analysis,including the shoulder angle,back angle,shoulder depth/width ratio and armpit depth/width ratio.Combined with these four important body parameters,the neck-shoulder shape of young women was classified following the established classification rules.Based on the front and side images,the parameters required for neck-shoulder shape classification were obtained by extracting the human body contour and identifying the feature points,and the automatic identification of the neck-shoulder shape was achieved according to the body type classification rules.The results show that young women′s neck-shoulder shape is divided into four categories,namely round wide shoulder,flat narrow shoulder,round drop shoulder,hunchback flat shoulder,accounting for 25.53%,23.94%,25.59%and 23.94%,respectively,of the total sample.The identification of the neck-shoulder shape based on the front and side images of 40 test samples is verified,and the accuracy ratio reaches 90%,indicating that the neck-shoulder shape automatic identification system constructed using this method is effective.

关 键 词：颈肩部形态 体型分类 图像 尺寸提取 自动识别 

分 类 号：TS941.17[轻工技术与工程—服装设计与工程]