基于三维测量的青年女性臀部体型概率神经网络识别模型构建  

Construction of recognition model for young females' hip shapes using probabilistic neural network( PNN) method based on 3-D body measurement

作　　者：金娟凤[1] 杨允出[1,2] 夏馨[1] 倪世明[1] 邹奉元[1,2] 

机构地区：[1]浙江理工大学服装学院,浙江杭州310018 [2]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》2014年第4期100-104,共5页Journal of Textile Research

基　　金：浙江省自然科学基金资助项目(Y1110504;LQ12E05015)

摘　　要：快速准确地实现体型识别是人体体型研究的热点。为满足服装臀部合体性的要求,结合青年女性臀部体型特征,构建了基于三维测量的青年女性臀部体型PNN识别模型。首先,运用三维人体测量仪采集数据,并提取6个典型指标,进行臀部体型细分;其次,引入概率神经网络方法,构建以典型指标作为输入层,体型类别作为输出层,径向基函数作为模式层的网络结构模型;再次,利用MatLab R2009a软件对构建的概率神经网络模型进行仿真实验,通过训练获取精度高,结果稳定的模型;最后,测试模型识别精度。结果表明,该模型识别率高,识别性能良好,为女性臀部体型识别提供了一种新方法,同时也拓宽了概率神经网络方法的应用领域。How to make a fast and precise recognition of body shapes has become a hot spot in the field of body shape research. In order to improve the fitting of garment on hips, the characteristic of young females＇ hip shapes was taken into consideration and a recognition model for young females＇ hip shapes using probabilistic neural network （PNN） method based on 3-D body measurement was constructed. At first, hip size data were acquired based on 3-D body measurement, of which, 6 typical indices were extracted, and hip shapes were classified; then probabilistic neural network （PNN） method was introduced, and a model, whose input layer contained six typical indices, output layer was shape types and pattern layer was radial basis function （ RBF）, was constructed; the simulation was completed by MatLab R2009a, and an exact and steady model was obtained by training; and finally, the accuracy of the model was tested. The result showed that the model had excellent performance with high prediction accuracy, so that it could provide a new method to recognize young females＇ hip shapes and widen applicability of PNN method as well.

关 键 词：三维测量 臀部体型 概率神经网络 识别模型 

分 类 号：TS941.17[轻工技术与工程—服装设计与工程]