基于面料力学属性的三维织物模拟

为了精确再现虚拟服装中面料的力学属性,提出一种模拟不同材质织物变形效果的方法.首先测试样本的力学性能参数,通过相关性分析获得其中与悬垂特性密切相关的部分参数,以此作为神经网络的输入数据;然后对比实拍样本及其静态仿真效果图,将获得样本仿真模型控制参数的期望值作为BP神经网络的输出数据;最后建立BP神经网络实现织物力学属性与仿真模型控制参数之间的非线性映射.应用文中方法对15种常用服装面料进行实验的结果表明,仿真效果较好地体现了面料的材质信息,在此基础上实现的虚拟服装更加精准,有助于提高系统的真实感.

3D Textile Simulation Based On Fabric Mechanical Properties

We present a new method to simulate the deformation of fabrics with different materials for the purpose of accurately reproducing the mechanical properties of fabrics in virtual clothing.Our method first measures the mechanical parameters of samples,and computes parameters that are closely related to the draping characteristics by performing correlation analysis.Thereafter,the real-captured images of samples are compared with the simulated images,yielding a set of control parameters of the simulation model.Finally,we employ a Brief Propagation based neural network algorithm to characterize the non-linear mapping between fabric mechanical properties and the control parameters of simulation model,whose input and output are the selected parameters of the correlation analysis and the control parameters of the simulation model.Experiments were carried out for 15 commonly used fabrics,and the results demonstrate that our method faithfully expresses the material information of fabrics,and helps improve the simulation realism.