【作者】 陈敏之；

【导师】 张渭源；

【作者基本信息】 东华大学 ， 服装设计与工程， 2012， 博士

【摘要】 近年来,服装电子化量身定制(Electronic Made to Measure)理念以广泛为人们所接受。其中三维人体模型的建立是实施EMTM的关键环节之一,其形态的优良性决定了定制服装最终的效果。而三维人体模型既要建立在对人体净体基本尺寸的正确描述上,又要再现人体穿着基础内衣(如文胸)后的塑型效果,为后期的服装定制与展示打下基础。本文旨在改变以往在单一状态下进行人体体型测量与分析的研究思路,而将净体特征与文胸着装后的胸部形态关联起来。由于人体胸部的变形能力以及文胸的结构设计对文胸的着装效果均有显著地影响,同时鉴于人体的体型特征是文胸的设计的基础,而文胸的款式复杂多样而且在不断的发展,因此在本次的研究中,通过采用固定文胸的结构来最大可能地去除塑型内衣对胸部特征产生特殊影响,重点研究不同人体体型胸部自身形态改变能力,以寻求对文胸着装下胸部塑型效果的预测和评估方法。具体的研究内容如下：1)文胸着装状态下女性胸部形态变化特征研究设计了基于三维测量技术的测试用基础文胸着装前后女体胸部形态效果对比实验。通过配对样本T检验,对19项与服装着装效果、服装样板制作密切相关的人体体型变化量参数进行筛选,在剔除4个t检验显著性概率P>0.005的体型变化变量后,最终确定15个形态变化量作为反映女体胸部形态变化的特征参数。运用因子分析,将胸部形态变化量划分为四组,分别用于反映胸部截面、挺度、丰度和聚胸等方面的形态变化效果。2)基于熵权的胸部形态变化效果评价模型研究本文将信息熵理论引入到对服装着装效果客观评价中。选取了经过筛选后的胸部形态变化特征参数Dm作为文胸着装形态变化效果评价指标基于文胸着装效果实验样本,通过建立评价指标体系、指标熵权计算、综合得分计算等过程,分别从截面形态、挺度效果、丰度效果、和聚胸效果等四个方面,建立基于熵权的综合评价模型,分别将胸部形态变化程度分为弱、适中、强3类,从理论上总共得到81个类别,其中实际存在的种类为37种。通过统计检验,分类模型具有较好的实施效率和分类的有效性。根据体型变化量熵权分配值以及其在服装制作中的重要性,选取胸围、前颈到胸围线直线距离、乳下长以及乳间距等四个体型变化量作为反映胸部形态变化程度的关键尺寸,并建立了关键尺寸与其他特征变化量的回归模型。3)构建基于改进型神经网络集成的胸部塑型能力预测模型通过净体特征数据与文胸着装状态下截面、挺度、丰度、聚胸等变化效果评价值的Spearman相关分析,提取了对胸部塑型影响较大的净体体型特征参数,并以此为基础设计和构建了胸部塑型能力评价预测模型。该模型采用了并行的四个神经网络集成模式分别对四项胸部塑型能力进行仿真预测。在基于对神经网络泛化能力及集成理论的分析,本文提出了基于聚类的成员网络训练集提取方法和动态权重分配法应用于神经网络集成的设计,提高了成员网络精度,增加了成员网络间的差异性,从而增强了神经网络的泛化能力。采用了剔除法来确定所适合的隐层结构,继而设定改进型BP网络模型结构和训练参数。通过对胸部塑型能力评价预测模型的仿真检验,证明该模型能基于个体的胸部净体数据,以基础文胸着装状态为参照标准,实现个体的胸部塑型能力的预测,特别是对截面形态、挺度效果、丰度效果等方面塑型能力的评价,准确率在85%以上。4)胸部塑型能力评价与预测系统设计与开发从女体的胸部形态特征分析在服装行业各个环节中的需求入手,对胸部塑型能力评价与预测系统在服装设计、生产与销售过程中的实用性和应用前景进行了分析。并基于胸部形态评价的熵权分析、胸部塑型能力预测神经网络集成、以及胸部形态变化分类描述等研究成果,编程开发了胸部塑型能力评价与预测可视化软件系统。系统通过交互式导入文胸着装前后的人体测量数据,可快速实现对胸部形态变化定性、定量的客观评价；亦可导入个性化人体净体胸部特征参数,以穿着测试用基础文胸的状态为参照,实现对胸部截面、挺度、丰度、聚胸等四个方面形态变化能力的预测,并对胸部、前颈点到胸围线距离、乳下长、乳间距等四个胸部形态变化关键尺寸的变化量进行定量的估计。该系统可直接运用三维测量数据,输出用于数字化服装设计制作和展示的胸部形态变化定性、定量数据,可实现与多种数字化测量与设计系统的有效链接,从而提高整个生产体系的实施效率。5)个性化三维虚拟胸部变化模型设计个性化的女体胸部形态三维模型是服装制作、三维试衣显示的基础。然而,由于女体胸部形态常受到文胸的影响,净体尺寸与文胸着装后人体尺寸相差较大,且不同的人体胸部形态变化程度不一样。本文通过三维净体人体扫描数据和NURBS曲面建模构建女体胸部净体三维模型,并创新性的提出了关键变化尺寸控制的三维胸部模型变化实施方案,以胸围线距离、乳下长、乳间距等四个胸部形态变化关键尺寸作为控制参数,利用SDD变形技术快速实现个性化的三维虚拟净体胸部模型向文胸着装后虚拟人体的转换。并通过实践证明,该方法能够快速有效的模拟文胸着装后的胸部形态特征。本课题探讨了对服装设计、制作和试装模拟起重要影响的女体胸部形态变化规律,从胸部塑型能力的角度对个性化人体进行分类和预测识别,并在此基础上对胸部形态参数化控制三维人体虚拟变形进行了探索。本课题成果填补在女体胸部形态变化分析及其参数化控制模拟领域的研究空白,具有较强的研究价值和实用意义。更多还原

【Abstract】 In recent years, the concept of Electronic Made to Measure (EMTM) is accecpted widely. And the three-dimensional human virtual modeling is one of the keys to the implementation of EMTM. And its shape determines the final effect of custom clothing. However, the three-dimensional human virtual model should not only be based on the correct description of naked human body measurement, but also the deformation effect of the body wearing underwear (such as bra). What is more, it would be the basis of clothing customization and vitual showing.Changing the research ideas from measuring and analysising the shape of human body in a single state, the pupose of the paper is to connect the characteristics of naked body to the breast shape in bra. Generally, the final effect in bra is affected by both of body shape and structcure of bra. However, bra always is designed based on body shape, and the styles are various and develping. So through applying a fixed base bra to experiment, the paper focuses on the breast deformable abilities of different kinds of body shape, to find the method to predict and evaluate the deformable effect of breast in bra. The main works and approach in the study can be outlined as follows:1) Deformable characteristics of female breast shape in braComparative experiments of female breast shape before and after wearing bra was designed based on three-dimensional measurement technique. By paired sample T test,19measurments of breast deformation were screened, which were closely related to the pattern making and the final effect of clothing. After screening out4measurements whose t-test significance probabilities P were larger than0.005,15measurments of breast deformation were selected as the characteristic parameters.By the factor anlaysis, they were divided into four groups, to reflect the bust section, stiffness, chubbiness and gathering of breast deformation respectively.2) Entropy-based evaluation model of breast deformationThis paper introduced the entropy theory to objectively evaluate the effect of breast deformation. The15characteristic parameters were selected to evaluate the effect of breast deformation. With the experimental samples, the evaluation index system was established. Through calculating entropy weights and composite scores, the comprehensive evaluation models were established based on the entropy for the breast deformation of cross-section shape, stiffness, chubbiness, and gather respectively. From each aspect, the effect of breast deformation was devided into3categories, weak, moderate and strong. Therefore a total of81categroies was achieved, of which there existed37categories. It was proved that evaluation models can be implemented efficiently, and the classification is effective. Four variables were extracted as the key parameters reflecting the breast deformation. And the linear regression models of the related the characteristic parameters of breast deformation were established finally.3) Prediction model of breast deformable ability based on improved neural network ensemble argrithmThrough the Spearman correlation analysis, the characteristic parameters of naked body shape were extracted, which were correlated significantly to the evaluation scores of4aspects of breast deformation.Then a prediction model of breast deformable ability was constructed with4parallel ensemble neural networks. In this paper, a new method based on K-means cluster analysis was put forwards for achieving the training subsets of component ANNs. Also, the outputs of component ANNs were combined via a Dymaic Weight Distribution Method. It was shown that this approach improved the ensemble generalization abitlity. The proper structure of hidden layer for each of component ANN was determined by Remove Method. Finally, the effects of breast deformation of test samples were simulated, through which the accuracy of model prediction was validated. Especially, the mean correct rates for predicting the deformable effect of cross-section shape, stiffness and chubbiness were above85%.4) Development of the evaluation and prediction software system of breast deformation in braUsing the visualization technology, the intelligent prediction software was programmed with the achievements in the entropy-based evaluation model of breast deformation, ensemble ANNs for breast deformable ability prediction, and the description of the classifications. This software could directly make the qualitative and quantitative evaluation of breast deformation objectively by loading the measurements of body shape before and after wearing bra. In addition, it could predict the4items of breast deformable ablitity for an individual, by inputting the characteristic parameters, and output the quantitative estimate of the four key parameters for the breast deformation in normal bra.This system could be connected with the digital systems used in EMTM, such as three-dimensional body scanner, or the virtual-try-on system. It would be helpful to improve the operational efficiency of EMTM.5) Development of the3-dimensional virtual model for individual breast deformation3-dimensional body shape model for an individual is the basis of pattern making and virtual try-on. However, the breast shape is easily to be affected by bra, which causes the large difference between the measurements before and after wearing bra. What is more, the deformation varies in different body shape. In this research, an innovative implementation was applied to virtualize the breast deformation by controlling the four key parameters. Firstly, the individual breast was modeled on the basis of3D scan data and surface modeling technology. Then SDD (Skeleton Driven Deformation) was applied to transfer an individual naked breast model to the deformated breat in bra, while the expected measurements of the key parmaters were satisfied. In conclusion, the breast deformation in bra was analysized in this research, which plays important roll in fashion design, pattern making and virtual-try-on. The individuals were classificated and evaluated by their breast deformable abilities. On the basis of it, the virtual breast deformation model by parametic controlling was explored. The achievement of this research filled up the blank of female breast deformation analysis and the related virtual simulation, with great value and pratical significance.更多还原