基于骨架结构的三维人体模型检索算法研究

发布时间：2019-04-17 10:09

【摘要】：三维模型已成为继声音、图像、视频之后的第四种数字化媒体技术。随着多媒体图像、模型采集设备和三维建模仿真技术的快速发展,三维模型数量出现爆炸性的增长。面对海量的三维模型,如何对其进行有效管理,从而实现高效便捷的三维模型检索和信息获取成为当前亟待解决的问题。三维模型骨架不仅能很好地表示模型的形状信息和拓扑骨架结构信息,还可以看成是对原始模型的一种压缩技术,大幅度地减少模型的存储空间,提高计算机存储空间利用率。随着图形学和可视化技术的发展,三维骨架成为三维模型检索的一个重要研究方向。目前,三维模型骨架已被广泛应用于计算机辅助制图、人体器官重建、虚拟现实、动画仿真、工业制造以及数字娱乐等领域。本文提出了一种创新性的人体三维模型检索算法。本文主要内容及创新点如下:1)数据库构建:从现有三维模型数据库中挑选部分形状结构较为复杂的三维模型,组成一个全新的数据库。使其更有利于三维模型检索算法的应用。2)骨架特征提取:采用基于L_1中值的三维模型骨架提取算法,并针对实际应用中原始模型的一些缺陷限制,提出了基于拓扑分裂的采样优化、基于密度权重更新的分布优化、基于椭圆拟合的位置优化等改进算法;3)匹配算法:提出了一种基于随机游走的超图匹配算法。该算法引入张量概念和随机游走的思想。通过不同阶张量表示不同阶超图边的信息,并将低阶张量以递归方式融合生成一个高阶张量,更有效的对超图结构进行约束,使得算法在特征点扭曲、外部变形、孤立噪声等情况下,有良好的匹配效果。通过与一些现有匹配算法进行对比,本文所提算法具有良好匹配检索性能,能够有效地检索出拓扑结构相似的模型。
[Abstract]:Three-dimensional model has become the fourth digital media technology after sound, image and video. With the rapid development of multimedia image, model acquisition equipment and 3D modeling and simulation technology, the number of 3D models increases explosively. In the face of massive 3D models, how to manage them effectively so as to achieve efficient and convenient 3D model retrieval and information acquisition has become an urgent problem to be solved. The 3D model skeleton can not only well represent the shape information and topological skeleton information of the model, but also can be regarded as a compression technique for the original model, which can greatly reduce the storage space of the model and improve the utilization rate of the computer storage space. With the development of graphics and visualization technology, 3D skeleton becomes an important research direction of 3D model retrieval. At present, three-dimensional model skeleton has been widely used in computer-aided cartography, human organ reconstruction, virtual reality, animation simulation, industrial manufacturing and digital entertainment. In this paper, an innovative 3D human model retrieval algorithm is proposed. The main contents and innovations of this paper are as follows: 1) Database construction: select some of the three-dimensional models with complex shape and structure from the existing three-dimensional model database to form a brand-new database. 2) Skeleton feature extraction: a three-dimensional model skeleton extraction algorithm based on L\-1 mean value is adopted, and some limitations of the original model in practical application are pointed out. Some improved algorithms such as sampling optimization based on topology splitting, distribution optimization based on density weight update, and location optimization based on ellipse fitting are proposed. 3) matching algorithm: a hypergraph matching algorithm based on random walk is proposed. This algorithm introduces the concept of Zhang Liang and the idea of random walk. By representing the information of different order hypergraph edges through different order tensors, the low-order tensors are fused recursively to generate a higher-order tensor, which more effectively constrains the hypergraph structure, so that the algorithm distorts and deforms externally at the feature points. In the case of isolated noise and so on, there is a good matching effect. Compared with some existing matching algorithms, the proposed algorithm has good matching and retrieval performance, and can effectively retrieve models with similar topological structure.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】