基于图像的人体肩部特征点检测方法  

Detection of human shoulder feature points based on image analysis

作　　者：邓中民[1] 王健恺 靳晓凝 魏宛彤 于东洋 柯薇[1] DENG Zhongmin;WANG Jiankai;JIN Xiaoning;WEI Wantong;YU Dongyang;KE Wei(State Key Laboratory of New Textile Materials and Advanced Processing Technology,Wuhan Textile University,Wuhan,Hubei430200,China;School of Fashion,Wuhan Textile University,Wuhan,Hubei430200,China)

机构地区：[1]武汉纺织大学省部共建纺织新材料与先进加工技术国家重点实验室,湖北武汉430200 [2]武汉纺织大学服装学院,湖北武汉430200

出　　处：《纺织学报》2023年第9期175-179,共5页Journal of Textile Research

基　　金：湖北省技术创新专项项目(2019AAA005)。

摘　　要：针对现有检测算法对肩部特征点的误检、漏检和检测误差大等问题,提出了一种改进的基于人体轮廓编码的肩部特征点的检测方法。该方法以人体所占图像的大小来确定特征链码串的长度,并且通过特征区域分割、特征区域遍历,具有几何特征的链码串动态筛选的搜索方式,筛选出符合条件的肩部特征点。在获取肩部特征点以后,计算出肩宽值,并且与实际人工测量的肩宽值进行比对与分析。本文实验共选取了100名不同身型的青年测试者(50男/50女)在同一拍摄环境下对该方法的可行性进行测试。结果表明,采用该方法能够快速识别人体的肩部特征点,与人工测量值对比得出的平均误差仅在3%左右,较好地提升了二维人体肩宽尺寸检测的效率,并且为后续服装匹配等相关工作提供了更加精确的数据支持。Objective Human body size data plays an important reference role in measuring human body type.Along with the continuous development of computer technology,non-contact anthropocentric methods based on two-dimensional images were reported to obtain human body size data,the accuracy of which depends largely on accuracy of the feature points of each part detected using these methods.For the shoulder,which is a feature part with fewer obvious feature point characteristics,the existing two-dimensional detection methods leads to large errors in the feature points,calling for further improvement.Method In order to address the problems of false detection,missed detection,and large detection errors of the existing detection algorithms for shoulder feature points,an improved detection method for shoulder feature points based on human contour coding was proposed.The method was capable of determining the length of the feature chain code string by the size of the image occupied by the human body and selecting the feature points of the shoulder by a search method of feature partitioning,feature region traversal,and dynamic screening of the chain code string with geometric features.After acquiring the shoulder feature points,the shoulder width values were calculated based on the location of the feature points and compared with the actual manually measured shoulder width values to analyze the experimental results.Results In this research,a total of 100 young testers of different body types(50 males/50 females)were selected to test the feasibility of the method under the sam e shooting environment.The experimental study showed that the method can quickly and accurately identify the shoulder feature points of the human body,and the locations of the selected feature points exist and are single,reducing the occurrence of false detection and omission(Tab.1).The average error derived was only about 3%,which verifies the feasibility of the method.Conclusion While maintaining the advantage of the low cost of the 2-D measurement method,it

关 键 词：肩部特征点 肩峰点 人体轮廓编码 人体尺寸测量 特征链码串 肩宽 

分 类 号：TS941.17[轻工技术与工程—服装设计与工程]