【作者】 刘华；

【导师】 鲍虎军；

【作者基本信息】 浙江大学 ， 计算机应用与技术， 2007， 博士

【摘要】 如今，随着计算机图形学技术的广泛应用，使得对三维数字化模型的需求日趋增长。三维数字化造型技术的应用已不再局限于传统的工业设计与制造、影视特效等领域，而昌逐渐渗透到人们的日常生活当中：从家庭娱乐、个人通讯、数字媒体，到日常办公等各个方面，都可以见到它的身影。由此可见，三维造型技术的普及与发展大有可为。然而，通用三维造型技术从操作的简易性、造型的效率、方法的普适性以及对用户领域知识的需求度等几方面考察，还未能真正满足技术普及化的需求。同时，对于分形、L-系统等多种特色造型技术，虽然可以高效地构造目标模型，但是从理论上仍然未能针对特殊领域系统地提出解决方案，在用户领域知识的需求度以及操作的简易性上仍存在诸多局限。针对这一现象，本文对特定应用领域的造型需求进行分析，引入软件工程中的“域工程”概念，提出了面向域的造型技术框架。区别于传统造型技术是造型工具以及造型资源的重用，面向域的造型技术更进一步，侧重于造型经验知识的重用。本文方法利用领域内的配置知识，包括模型的缺省配置、构造规则、非法组合和约束等信息来辅助设计者造型。这样，不但能够提高造型效率，一定程度上保证模型生成质量，并且降低了用户的专业知识需求。除此之外，文章还就了技术框架中的关键算法和步骤进行了详细的讨论。总体上说本文的贡献如下：●从面向域的问题求解角度，将软件工程中的“域工程”思想引入几何造型，提出了面向域的造型技术框架。●在领域内造型组件的获取方面，提出了一种新的基于散乱点的网格重建方法。利用图像的连续性和适当的二维交互来确保重建网格的拓扑正确性。采用递进式的网格优化过程，通过网格重整技术保证了网格质量，从而可以获得一系列精度下的高质量网格模型。●在面向域的造型描述方面，提出了一种新的建筑造型语法—构造语法，并应用于中国古代建筑的快速造型。用户只需输入几个简单的特征与风格控制参数，系统即可快速地自动生成所需的建筑单体和群落。●在造型控制方面，提出了规则辅助的半自动造型方式，该方法将领域内的配置知识通过辅助规则的形式加以描述。在造型的过程中，通过配制知识的约束，用户的交互量被极小化，从而提高了造型的效率。总之，本文所提出的面向域的造型技术框架实际上是一个开放性的系统结构。它的核心就在于利用领域内的配置知识来辅助用户造型。对于领域知识的描述、造型组件，以及造型的控制方法在本文中提出了一些可行的解决方案，但并不局限于此。许多现有的相关技术都可以被整合于这一框架之下。本文的研究成果可被应用于面向普通用户、面向特定领域的三维造型系统的开发。支持工业造型设计、数字娱乐等应用，通过预定义的领域内知识的辅助，所开发的系统可帮助用户更加高效和方便地构造出所需三维模型。更多还原

【Abstract】 Nowdays, as the techniques of computer graphic are widely exploited, more andmore 3D models are required in various domains, such as traditional industry, advertising,digital entertainment etc. Unfortunately, 3D modeling is a hard and professional job,which does not suit for most of the non-professional users. Here, we will discuss the 3Dmodeling techniques in domains.First, we review the history of modeling techniques in the last forty years and classifythem into three classes based on their operating schema. After that, we discuss the char-acteristic and applicability each of them. Based on the analysis, the purpose and contentof our research are presented.Secondly, we introduce the concept of "domain engineering" from the software en-gineering domain. We treat the 3D modeling problem in a different way and represent anovel framework of rapid modeling in domains. Not only the modeling tools and modelresources are reused, but also the experience of modeling is reused too. Configure knowl-edge which includes default configuration, construction rules, illegal combination andconstrains is stored in the domain engineering. All these knowledge can be used to as-sist users during the modeling process. They help to improve the modeling efficiencygreatly and also guarantee the quality of result models. After introducing the framework,we discuss three key problems of it, which are components(the basic unit of 3d models),the language of domains(how to represent the configure knowledge) and the control ofmodeling process.When discussing components, we focus on the problem of surface reconstructionfrom unoriented points, which is a easy way to acquire components. A practical approachof surface reconstruction is represented, which is different from the origins. Instead ofestimating local surface orientation, we first generate a set of depth images from the in-put point clouds, and a coarse mesh is then generated based on them by space carvingtechniques. The result mesh is progressively refined by local mesh refinement and opti-mization according to surface distance measure. Our approach is easy to implement, buthas the ability to outputs high quality meshes in different resolutions.We use semantic rules as the domain language. After analyzing the existing proce-dure modeling technology, we propose a new modeling grammar-Construction Gram- mar. It is more powerful than the previous ones and we experience it on the modeling ofChinese ancient architectures. An Chinese ancient architecture with thousands of facesmay be built in a few seconds. All the generated models are in strict with the ancientChinese architectural style features.Finally, we discuss the methods of modeling control. An semiautomatice rules assistmodeling approach is proposed in the chapter five. Configure knowledge in the domainengineering, which includes constructive rules, default settings, constrains etc., are allpresented as assist rules. It includes construction rules, configure rules and constrain rules.All the assist rules are predefined and used to help users to reduce modeling operations.Different from the traditional approaches, our method not only reuse the modeling toolsand resources, but also reuse the experience of modeling.In conclusion, the framework of modeling techniques in domains is an open struc-ture. Many techniques not only the modeling techniques can be applied in this framework.In our thesis, we just present several methods to handle modeling problems in domains.The framework which we described in our thesis can be applied in the development of 3dmodeling system. By all the configure knowledge predefined in the domain engineering,the modeling efficiency and quality will be improved.更多还原