Novel methods for reflective symmetry detection in scanned 3D models
Novel methods for reflective symmetry detection in scanned 3D models
复制标题
扫描 3D 模型中反射对称性检测的新方法
DOI：
10.1109/ivcnz.2015.7761555
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
R. Green
中科院分区：
文献类型：
--
作者：
M. Stephenson;Adrian Clark;R. Green
在过去的十年中，检测3D模型中的对称性的概念已经得到了广泛的研究。已经提出了许多算法来识别3D网格内的反射对称性，并提取模型的对称性水平的定量测量。许多早期的工作集中在识别无噪声的3D模型中的对称性，大多数现有的方法无法有效地处理被噪声扭曲的模型，例如通常在扫描真实的世界中的对象时获得的模型。这份报告详细介绍了两个强大的和快速的算法，可用于各种各样的模型，以确定全球近似反射对称的设计和实施。这些方法还能够识别已被噪声扭曲或包含微小缺陷的模型中可能的对称平面，使其成为真实的世界对象的扫描模型的理想选择。假设平面由主成分分析确定，之后，所提出的算法给每个平面一个数值，对应于它的可能性是一个平面的全球近似反射对称。第一种算法使用顶点之间的Hausdorff距离来估计对称性，而第二种算法使用基于光线投射的方法。
The concept of detecting symmetry within 3D models has received an extensive amount of research within the past decade. Numerous algorithms have been proposed to identify reflective symmetry within 3D meshes and to extract a quantitative measure for the model's level of symmetry. Much of the early work focuses on identifying symmetry in noiseless 3D models with most existing methods unable to work effectively on models distorted by noise, such as those commonly obtained when scanning objects in the real world. This report details the design and implementation of two robust and fast algorithms, which can be used on a wide variety of models to identify global approximate reflective symmetry. These methods are also able to identify likely planes of symmetry in models that have been distorted with noise or contain minor imperfections, making them ideal for scanned models of real world objects. The hypothesis planes are determined by principal component analysis, after which the proposed algorithms give each plane a numerical value corresponding to its likelihood of being a plane of global approximate reflective symmetry. The first algorithm uses the Hausdorff distance between vertices to estimate symmetry, whilst the second uses an approach based on ray casting.