基于神经网络集成的女下装号型归档模型构建  

Construction of female’s bottom size grading model based on neural network ensemble

作　　者：孙洁[1] 金娟凤[1] 倪世明[1] 叶玲[1] 刘焘[2] 邹奉元[1,3] 

机构地区：[1]浙江理工大学服装学院,浙江杭州310018 [2]嘉兴学院设计学院,浙江嘉兴314000 [3]浙江理工大学浙江省服装工程技术研究中心,浙江杭州310018

出　　处：《纺织学报》2013年第8期110-114,共5页Journal of Textile Research

基　　金：国家国际科技合作专项专目(2011DF1351570);浙江省新苗人才计划孵化项目(2012R406069)

摘　　要：如何实现大样本的快速、精确号型归档是服装数字化批量定制系统构建的关键问题。结合下装结构设计的关键部位尺寸,在国家标准基础上将下装控制部位扩展为10项;以女裤装为例,分别构建身高和腰围的简单BP神经网络归档模型,通过平均影响值(MIV)筛选出对归档结果影响较大的5个控制部位,并将其作为集成归档模型的输入层,其中身高、腰围、臀围、腰围高、后裆长为身高归档模型的输入变量,腰围、臀围、大腿围、腰长、身高为腰围归档模型的输入变量;通过Adaboost算法集成10个简单BP神经网络,获得具有高精度和强泛化能力的女裤装号型集成归档模型。How to achieve fast and precise size-grading for large numbers of garment is a key issue to the construction of digitalized mass customization system.Combining with key measurement in designing the pivotal parts of women pants or bottoms,this paper expanded the control parts of bottoms into 10 items based on the national standard.Taking female trousers as an example,after constructing a simple size grading model with BP neutral network,the five control parts that have a major impact on the result of grading were selected based on mean impact value（MIV） and then they were used as the input layer of garment size grading model.In the model,height,waist circumference,hip circumference,waist height and back gore length worked as input variables of the height model,and waist circumference,hip circumference,thigh circumference,waist length and height used as input variables of the waist circumference model.Then,after ensembing 10 simple BP neural networks through Adaboost algorithm,it achieved the size-grading model with high precision and strong generalization ability.

关 键 词：批量定制 平均影响值 号型归档 神经网络集成 

分 类 号：TS941.17[轻工技术与工程—服装设计与工程]