基于LVQ神经网络的青年女性胸部识别模型构建  

Establishment of Recognition Model for Young Females'Breast Shapes Based on LVQ Neural Network

作　　者：沙莎 李诗怡 迟诚 万亚如 江学为 SHA Sha;LI Shiyi;CHI Cheng;WAN Yaru;JIANG Xuewei(Wuhan Textile Institute of Design Innovation and Fiber Science,Wuhan 430073,China;Wuhan Textile School of Fashion Design,Wuhan Textile University,Wuhan 430073,China;Wuhan Textile and Apparel Digital Engineering Technology Research Center,Wuhan 430073,China)

机构地区：[1]武汉纺织大学设计创新与纤维科学研究院,武汉430073 [2]武汉纺织大学服装学院,武汉430073 [3]武汉纺织大学武汉纺织服装数字化工程技术研究中心,武汉430073

出　　处：《纺织工程学报》2024年第1期69-79,共11页JOURNAL OF ADVANCED TEXTILE ENGINEERING

基　　金：国家自然科学基金项目(61802285);湖北省教育厅科学研究计划重点项目(D20201704);湖北省服装信息化工程技术研究中心开放基金项目(184084006)。

摘　　要：为提高青年女性胸部形态分类的准确率,填补文胸号型分类体系存在的缺陷,结合青年女性胸部体型特征构建了一种基于LVQ神经网络的青年女性胸部识别模型。研究运用非接触式激光三维技术共采集216个青年女大学生胸部数据,将因子分析提取的9个胸部特征指标采用K-means聚类法,通过手肘图、轮廓系数图确定K值,最终将胸型分为4类。在此基础上构建LVQ神经网络胸型识别模型,以9项胸部特征指标为输入,4种胸型为输出,进行LVQ神经网络的训练。研究结果表明:模型经训练及测试后,识别精度达到95%,Kappa系数为0.932。与BP、PNN神经网络模型相比,在运算效率、模型精度和稳定性方面,LVQ神经网络模型的表现要明显优于其他两种神经网络。In order to improve the accuracy of young women's chest morphology classification and fill the defects of the bra number classification system,this paper constructs a young women's chest recognition model based on LVQ neural network by combining the young women's chest body shape features.The study uses noncontact laser 3D technology to collect a total of 216 young female college students'chest data,and the nine chest feature indicators extracted by factor analysis are clustered by K-means clustering method,and the K value is determined by elbow diagram and contour coefficient diagram,and the chest type is finally classified into four categories.On this basis,the LVQ neural network chest type recognition model was constructed,and the LVQ neural network was trained with the 9 chest feature indicators as input and 4 chest types as output.The results show that the model is trained and tested with a recognition accuracy of 95%and a Kappa coefficient of 0.932.Compared with the BP and PNN neural network models,the LVQ neural network model significantly outperforms the other two neural networks in terms of computational efficiency,model accuracy and stability.

关 键 词：三维人体测量 胸部特征 胸部形态分类 胸型识别 LVQ神经网络 

分 类 号：TS941.17[轻工技术与工程—服装设计与工程]