高质量动态四边网格序列生成算法研究

发布时间：2019-02-16 15:30

【摘要】：近年来,随着信息技术的不断发展,大量的三维扫描和重建技术不断涌现,人们已有能力获取三维模型的实时动态数据。将这些技术应用在计算机游戏和动画领域,可以比较方便地获得由系列三角网格模型表示的动画角色实时动作序列。但考虑到在动画的后期制作过程中,四边网格表示较之三角网格具有更适于动画编辑、贴图处理等优点,因而在数字游戏和动画电影制作流程中被广泛使用。但由于四边网格同时具有的全局与局部特点,四边网格的生成过程既要满足局部的几何限制要求又要满足全局性的需求,因此生成满足要求的高质量四边网格是计算机图形和动画领域研究的一个热点和难点问题。本学位论文以三维动画角色模型的动态三角网格序列为输入,对动态四边网格序列的生成和优化方法进行了研究。在调研数字动画和电影制作流程中对动画模型的实际需求基础上,主要研究了如何避免三维模型动画形变过程中的失真问题、如何表达动画模型常见的对称问题以及如何对动画三维模型全局优化问题。分别提出了支持多分辨率的具有动画感知的四边网格序列生成算法、具有对称属性的三维模型动态四边网格生成算法以及全局优化的四边网格序列自动生成算法,并取得了一系列的研究成果,在此基础上构建了一个三维角色模型四边网格序列优化重建几何处理框架。本文的主要研究工作和贡献如下:1、设计实现了自适应多分辨率动画感知四边网格序列生成算法。本文通过对形变拉伸系数的分析,提取出表征形变密度的交叉场的长度场,用来表示四边网格的分辨率需求,并在此基础上进行全局参数化,最终生成满足多分辨率需求的高质量四边网格序列。与以往方法相比,本文方法更好地解决了三维动画模型在形变过程中的形变失真问题,生成的每一帧四边网格在表达精度上、四边网格的正交性等方面均效果更优,更加适用于形变较大的三维角色模型的网格重建,更加符合动画建模师的要求。2、设计实现了对称模型的四边网格序列生成算法。针对计算机游戏动画中较高频率出现的人或动物等形状对称模型,研究实现一种简单交互的对称检测方法,方法通过计算权衡表达对称属性和形变属性的交叉场,生成的四边网格序列与已有方法相比,不仅在每一帧都能更好的表达全局对称特点和局部几何形变特点,且奇异点更少。3、设计实现了全局优化的四边网格序列生成算法。与已有针对单帧或者将整个动画序列作为一个整体来研究的工作不同,本文采用分组策略将输入的动画序列进行分解得到多个子序列,通过计算子序列交叉场进而对子序列交叉场进行优化合并,使得最终生成的交叉场能够更好地表达动画序列的几何和拓扑结构,进而生成高质量的四边网格序列。
[Abstract]:In recent years, with the continuous development of information technology, a large number of 3D scanning and reconstruction techniques have been emerging, people have the ability to obtain real-time dynamic data of 3D models. By applying these techniques to computer games and animation, real-time action sequences of animated characters represented by a series of triangular mesh models can be easily obtained. However, in the post-production process of animation, quadripartite grid representation is more suitable for animation editing and mapping than triangular grid, so it is widely used in digital game and animation film production process. However, due to the global and local characteristics of quadrilateral meshes, the generation of quadrilateral meshes should not only meet the local geometric constraints but also satisfy the global requirements. Therefore, the generation of high quality quadrilateral grids is a hot and difficult problem in the field of computer graphics and animation. In this dissertation, the generation and optimization of dynamic quadrilateral mesh sequences are studied with the input of dynamic triangular mesh sequences of 3D animation role models. On the basis of investigating the actual demand of animation model in digital animation and film making process, this paper mainly studies how to avoid the distortion in the process of 3D model animation deformation. How to express the common symmetry problem of animation model and how to optimize the animation 3D model globally. An algorithm for generating multi-resolution quadrilateral mesh sequences with animation awareness, a dynamic quadrilateral mesh generation algorithm for 3D models with symmetric attributes and a global optimization algorithm for generating quadripartite mesh sequences are proposed respectively. A series of research results are obtained, and a geometric processing framework of 3D role model based on quadrilateral mesh sequence optimization is constructed. The main research work and contributions are as follows: 1. Adaptive multi-resolution animation sensing quadrilateral mesh sequence generation algorithm is designed and implemented. In this paper, the length field representing the deformation density is extracted by analyzing the deformation stretching coefficient, which is used to represent the resolution requirement of the quadrilateral grid, and then the global parameterization is carried out. Finally, a high quality quadrilateral mesh sequence is generated to meet the multi-resolution requirements. Compared with the previous methods, the method in this paper solves the distortion problem of 3D animation model in the process of deformation. The generated quadrilateral mesh is more effective in terms of the accuracy of representation and the orthogonality of the quadrilateral mesh. It is more suitable for mesh reconstruction of 3D role models with large deformation and meets the requirements of animation modelers. 2. An algorithm for generating quadrilateral mesh sequences of symmetric models is designed and implemented. In this paper, a simple interactive symmetry detection method is developed for the symmetry model of human or animal, which appears in high frequency in computer game animation. The method is used to calculate and balance the cross-fields that express symmetric attributes and deformed attributes. Compared with the existing methods, the generated quadrilateral mesh sequences can not only better express the global symmetry and local geometric deformation characteristics in each frame, but also have fewer singular points. 3. A globally optimized algorithm for generating quadrilateral mesh sequences is designed and implemented. Different from the previous research work on single frame or the whole animation sequence as a whole, this paper uses group strategy to decompose the input animation sequence to obtain multiple sub-sequences. By calculating the crossover fields of subsequences and combining them optimally, the resulting crossover fields can better express the geometric and topological structure of animation sequences, and then generate high-quality quadripartite mesh sequences.
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP301.6

【相似文献】