摘要
织物模拟技术利用计算机建立纱线三维模型，处理纱线纹理数据，根据织物编织模式，再结合计算机图形绘制技术，表现织物的真实感视觉效果。织物模拟在国内外已有较长时间研究，出现了相当多的成果，应用系统也陆续出现，其各项性能已接近达到实际应用水平。
本文针对提花织物特有的工艺、组织，综合国内外现有的织物模拟技术，提出若干实现方案并分析其优缺点，实现复杂度等因素；结合实验室和个人现有的研究水平，针对其中一种方案及其纹理问题，进行实际研究和开发，展示目前我们取得的成果，指出方案中存在的问题和我们将来的工作。
本实现方案首先建立纱线模型，分析输入纹织CAD文件中的组织信息，结合图形绘制技术输出效果图。纱线模型可以从输入的纱线数码图像中分析其形状特征参数来建立，但通常我们凭直观感觉建立数学表达式，观察生成的结果来调整表达式的参数，甚至建立更完善的表达式；同时我们运用数码图像中的纹理特征来简化对纱线复杂微观结构的建模，而又能满足表现纱线的真实感效果；结合模型和纹理，运用纹理映射技术，我们生成逼真的纱线。
本实现方案涉及到纹理分析、综合，纹理映射，提花织物三维组织分析等技术，论文主要对与纹理相关的技术作一定的讨论，纹理分析运用频谱、小波分析或马可夫随机场建模纹理图像在空间上的分布特征，纹理基元的种类和其分布。纹理综合在输入一小块的纹理基础上生成任意大小的纹理，当我们仔细分辨这两块纹理时，感觉它们是相同的。纹理映射把纹理图像映射到三维物体表面上，在物体各表面上纹理不会变形和间断，而且有很好的分辨率。
We use computer to model 3-D yams, process texture image data and produce image of visual effect by computer graphics, according the knitting pattern, in simulation of fabric. It's been several years we researched in this project, both internal and external. Many techniques have obtained and some application systems emerged. Most of functions are approaching practice.
In this paper we first suggests some schemes along with specific of jacquard fabric combined with latest graphics and vision techniques and compares their functions, advantage, complexities and so on. We worked on one of these schemes at the consideration of our research level. Later we exhibit our results, indicate the existed problems and our future work.
In this scheme we model yarns first, then analyze the knitting pattern from input CAD file, render the yams in fabric and map texture onto them. Though it is possible to model a yarn by extracting shape feature from digitalized yam image, we usually model it by experience. We adjust the parameters of the model by checking the result yam, even reformulate this model if not satisfied. We use the texture in the yarn image to simplify the modeling of complicated microstructure of yarn and this method could aso produce good result. Combined with the model and textue, we get visual yarns.
This scheme covers techniques of texture analysis, synthesis, texture mapping and 3-D knitting pattern analysis of fabric. We discuss here the techniques related to texture. Texture analysis uses frequend-domain, wavelet or MRF to get feature distribution of texture image, classes of texel and the rule of placement. Texture synthesis generates unlimited size texture image by input a patch of small texture, though sufficiently different, when perceived by human being, they are same under stochastic process. Texture mapping technique maps texture onto surface of 3-D object, generating no interruption anywhere and having fine resolution.
引文
[1.1] S.Z.Li. 《Markov Random Field Modeling in Computer Vision》http://www.vision.ee.ethz.ch/～rpaget/Markov/book.html。
[1.2] 彭群生，鲍虎军，金小刚编著．《计算机真实感图形的算法基础》．科学出版社，1999．
[1.3] 边肇祺，张学工等编著．《模式识别》第二版．清华大学出版社，1999．
[1.4] 章毓晋编著．《图像处理和分析》．清华大学出版社，1999．
[1.5] Haralick R M. Statistical and structural approaches to texture. Proc IEEE,1979, 67:786～804
[1.6] Gool L V, Dewaele P, Ostredinck. Texture analysis anno 1983. Computer Vision Graphics Image Processing, 1985, 29:336～357
[1.7] Reed T R, DuBuf J M H. A review of recent texture segmentation and feature extraction techniques. Image Understanding, 1993, 57:359～372