三维模型整体与局部相似性评价方法研究

发布时间：2018-11-10 12:55

【摘要】：现代产品研发设计中广泛采用自顶向下、由粗到精、由整体到局部的设计方式,但由于结构非常简单的零件也需要花费相当长的时间进行功能、结构、工艺等方面的重新设计,如果能将现有产品或技术通过适当的结构修改等来满足新功能要求,则在保证产品质量的同时,可以有效降低成本,强化零部件之间、装配体之间、甚至产品之间的继承关系。因此,设计重用对于企业来讲具备重大战略意义。三维模型因其包含多层次的信息集合、丰富的视觉细节而得到越加广泛的应用:基于三维数字化模型的产品全生命周期管理是现代产品设计制造的重要发展趋势;同时,企业在生产过程中积累了大量的、丰富的三维模型,研究如何从产品海量CAD模型中快速、有效地查找出所需模型,充分挖掘和发现其蕴含的信息和知识并对其加以重用,是加快产品开发进程、缩短产品制造响应周期、提高产品研制质量的重要途径之一。农业机械由于交互对象的复杂性等研制过程同一般机械产品相比困难较多,将较为前沿的设计方法逐步应用到农业机械中对于提高我国整体农业机械化水平具有无可替代的重要意义;以文本关键字为基础的文本匹配方式由于其自身固有的描述性缺陷在模型海量增长情况下已不能适应新时代下对三维模型的重用要求。基于此,基于内容的三维模型自动检索技术在农机设计重用中具备越加重要的研究价值。本文在总结国内外三维模型检索的研究思路、层次、技术之后,重点进行了以下三方面研究:(1)一种基于距离面积几何特征的单个模型整体相似性评价方法。基于统计特征的三维模型整体搜索方法因其原理易于理解,算法效果良好得到了广泛的研究。本文算法通过对经典形状分布算法进行补充性改进,优化随机取点方式后减少采样点数量并增加模型面与面之间的面积比值特征,在提高复杂模型分辨率的同时能较大地提升算法时间性能;最后利用离线特征提取技术与在线模型匹配技术,进行三维CAD模型整体相似性评价。(2)一种基于模拟退火算法的模型局部结构挖掘方法。考虑到大多通用评价方法不适用于模型典型局部结构,本文对模拟退火方法在模型局部结构中的挖掘能力进行验证。算法对模型几何拓扑属性进行总结研究之后,以属性邻接图为基础,构建相应邻接矩阵,改变搜寻思路,将在大图中寻找小图的问题转化为使用模拟退火算法对模型关联属性图的最大团进行挖掘,最终以蚁群以及遗传算法作为对比,试验证明算法可以作为初步的启发式局部结构挖掘方法。(3)针对以上两种算法,以C++作为算法开发语言,以Visual Studio.NET2008为集成开发环境,Open Cascade为几何造型平台,以部分农机专业模型与通用模型为试验对象,开发了基于Windows程序风格的三维数字化模型检索系统,便于进行设计重用。
[Abstract]:In modern product development and design, the top-down, coarse to fine and integral to local design methods are widely used. However, because the parts with very simple structure also take quite a long time to redesign in the aspects of function, structure, process, etc. If the existing product or technology can meet the new functional requirements through appropriate structural modification, it can effectively reduce the cost and strengthen the inheritance relationship between parts, assemblies and even products while ensuring the product quality. Therefore, design reuse is of great strategic significance to enterprises. 3D model is more and more widely used because of its multi-level information set and rich visual details. The product lifecycle management based on 3D digital model is an important development trend of modern product design and manufacture. At the same time, the enterprise has accumulated a large number of rich three-dimensional models in the process of production. This paper studies how to quickly and effectively find the required model from the massive product CAD model, fully mining and discovering the information and knowledge contained in it and reusing it. It is one of the important ways to speed up the process of product development, shorten the response period of product manufacture and improve the quality of product development. Because of the complexity of interactive objects, the development process of agricultural machinery is more difficult than that of general mechanical products. The application of more advanced design methods to agricultural machinery is of irreplaceable significance for improving the overall level of agricultural mechanization in China. Because of its inherent descriptive defects, the text matching method based on text keywords can not meet the requirements of reusing 3D models in the new era under the condition of massive growth of models. Based on this, the automatic retrieval technology of 3D model based on content has more and more important research value in the reuse of agricultural machinery design. After summarizing the research ideas, levels and techniques of 3D model retrieval at home and abroad, this paper focuses on the following three aspects: (1) A method for evaluating the global similarity of a single model based on geometric features of distance area. The global search method of 3D model based on statistical features is easy to understand and the algorithm has been widely studied. In this paper, the classical shape distribution algorithm is improved to reduce the number of sampling points and increase the area ratio between the model surface and the surface. The time performance of the algorithm can be greatly improved while improving the resolution of the complex model. Finally, the global similarity of 3D CAD model is evaluated by off-line feature extraction and on-line model matching. (2) A method of local structure mining based on simulated annealing algorithm is proposed. Considering that most general evaluation methods are not suitable for typical local structures of models, this paper verifies the mining ability of simulated annealing methods in local structures of models. After summarizing and studying the geometric and topological attributes of the model, the algorithm constructs the corresponding adjacency matrix based on the adjacent graph of the attributes, and changes the search thinking. The problem of finding a small graph in a large graph is transformed into mining the largest cluster of associated attributes in the model using simulated annealing algorithm, and finally taking ant colony and genetic algorithm as the comparison. Experiments show that the algorithm can be used as a preliminary heuristic local structure mining method. (3) for the above two algorithms, C is taken as the algorithm development language and Visual Studio.NET2008 as the integrated development environment, Open Cascade as the geometric modeling platform. A 3D digitized model retrieval system based on Windows program style is developed with part of agricultural machinery professional model and general model as experimental object, which is convenient for design reuse.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S220;TP391.7

【参考文献】