【作者】 刘正；

【导师】 陆国栋；

【作者基本信息】 浙江大学 ， 机械设计及理论， 2013， 博士

【摘要】 伴随着时尚产业快速发展和个性化潮流的到来,人们对服装的要求逐渐从舒适、美观的大众化转变为修饰自我、彰显气质的个性化。服装个性化包含款式、合体、风格等多个层面偏好设计。传统服装CAD技术自底而上的设计流程和功能模块相对独立的设置,隔断了不同设计层面内容相互关联,无法反映服装要素组合形成特征过程中的构成规律,难以把握服装整体设计状态。为了体现服装个性化特征,需准确选择服装要素及组构形式,实现不同层面设计特征的关联呼应。因此,本文在分析服装构成要素基础上,提出了以要素耦合驱动来表征服装多个层面的偏好设计。通过人体参数与间隙量耦合形成着装空间类要素,由用户偏好驱动间隙量空间分布实现个性服装合体设计；通过造型特征与风格特征耦合生成造型风格类要素,由用户交互驱动造型实现个性服装进化设计；通过对服装耦合状态分析及干涉冲突处理,提出属性相似和用户评价相似的个性服装系列化设计。主要内容包括：提出基于着装空间的个性服装合体设计。基于因子分析和服装知识确定人体特征参数,利用特征参数驱动人体高维尺寸参数重构；通过三维测量试验获取着装间隙量数据,定义间隙量空间分布表述,采用数据挖掘技术建立了人体参数与间隙量分布的耦合关联；将三维间隙量数据转化为二维尺寸从而改进服装纸样设计,通过结合用户合体偏好和试穿模糊评价确定间隙量分布权重,利用权重调整人体特征部位间隙量数据驱动服装纸样修正。提出基于造型风格的个性服装进化设计。将服装造型要素解构为款式、装饰、色彩和图纹,明晰服装造型在图像信息层面的细节构成；采用感性工程获取服装整体意象风格特征,通过核主成分分析提取服装造型细节特征,利用支持向量机建立造型细节特征与整体风格特征的关联；在造型风格耦合下构建用户偏好驱动的遗传进化设计,将进化后代造型的风格关联计算值作为风格适应度,将用户交互评价作为偏好适应度,两者结合指导遗传进化过程,实现个性服装进化设计。提出基于多要素耦合的个性服装系列化设计。定义服装要素耦合状态分析方法,采用特征协同优化处理要素耦合干涉；定义服装系列耦合冲突检测方法,利用服装属性相似度计算处理服装间冲突,实现服装本征系列设计；引入协同过滤算法对服装协同评价,基于用户偏好要素改进近邻服装相似计算方法,基于服装多要素耦合特点,结合属性相似和评价相似两个层面获得相似服装集,基于提取要素特征进行服装集分类,实现服装主题系列设计。以本文研究成果为核心,结合采集人体数据,完成合体纸样设计实例,开发个性服装进化设计和系列化设计原型平台,展示进化设计和系列化设计实例。最后,总结全文研究内容,针对研究工作存在的不足提出未来研究的展望。更多还原

【Abstract】 With the booming of fashion industry and the trending of individuation, people’s requirements for clothing have been changed from comfort and beauty to individuation and temperament, and the individuation of garment involves pattern, fit and style. Traditional garment CAD design workflow is from underlying factors to apparel products and each function module of garment CAD is mutually independent, separating the correlative relationship of each design stage. This kind of workflow cannot reflect the combination laws of garment element, so that users cannot grasp the fashion design status.In order to reflect individual features of garment, the clothing factors and constituting form must be chosen accurately. This paper presents the design philosophy of driven elements coupling based on the analysis of the components of clothing. First, we acquire user preference through the elements of dressing space formed by coupling the human parameters and intervals, and realize personalized garment fit design by the spatial distribution of intervals driven by user preference. Second, we make up modeling style elements by coupling the styling features and style characteristics, and drive individual garment evolutionary design based on user interaction. Third, we deal with interference through analyzing the garment coupling status, and then propose a series design for personalized clothing combining two kinds of similarity calculations based on attribute similarity and user evaluation similarity. The main contents are as follows.Initially, we present individual garment fit design based on dressing space. The feature sizes are extracted by factor analysis and garment knowledge. The body detailed sizes are rebuilt using the feature sizes. By acquiring the data from the3D body scanner, the spatial ease allowance is defined. The coupling relationship between body sizes and ease allowance is built based on data mining. The spatial ease allowance can be mapped into2D ease allowance to modify the pattern design. Therefore, once the user fit preference and fuzzy fitting semantic evaluation are transformed into spatial ease allowance weights, the pattern can be adjusted by the spatial ease allowance which contains preference and evaluation. Subsequently, we propose individual garment evolutionary design based on styling features. The construct of garment is made clear through the detailed design content which is divided into style, ornament, color and pattern. Using Kansei engineering, the whole style features can be acquired, and the detailed style feature can be obtained by kernel principal component analysis (KPCA). Then, the relationship between two features can be established based on support vector machine. In order to ensure the stable feature of the evolutionary consequences, the style feature estimation is taken as a judgment in the genetic algorithm. In order to reflect the user preference, the interactive evaluation is adopted in the genetic algorithm as well. These direct the evolutionary design, improving the accuracy and efficiency of the garment genetic algorithm.At last, we raise individual garment series design based on multi-factors coupling. Through the analysis of the garment factors coupling status, the method deals with factor interference is given base on collaborative optimization. Aiming at the conflict between series garments, the attribute similarity method is proposed to deal with the conflict, and the series design for case characteristic can be achieved. The collaborative filtering algorithm is introduced in series design. By clustering the similar review, the similar garment can be searched, and then the similar garment set can be collected based on similar attributes and reviews. This set is classified into several themes according to the features extracted from the set.更多还原