A Mesh-Database-Independent Edge- and Face-Swapping Tool
A Mesh-Database-Independent Edge- and Face-Swapping Tool
复制标题
独立于网格数据库的边缘和面交换工具
DOI：
--
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
C. Ollivier
中科院分区：
文献类型：
--
作者：
C. Ollivier
软件复用是科学计算软件开发中越来越普遍的实践。最近，人们开始强调软件组件的定义：使用明确定义的接口来封装特定功能的软件对象。特别是，至少有两个这样的组件定义已经开发了低级别的网格功能，其中至少有一个已经有几个工作实现。这些组件为网格操作服务的开发打开了大门-例如交换，平滑，边折叠，面分裂等-这些服务独立于底层网格数据库的内部;反过来，更高级别的操作可以构建在这些之上。本文介绍了一个简单的网格交换在二维和三维的软件组件。应用可以控制用作交换标准的小区质量测量，包括定义新的质量测量。交换组件使用Terascale Simulation Tools and Technologies（TSTT）网格组件进行编码，因此可用于实现TSTT网格接口的任何网格数据库。结果给出了比较新的软件组件与特定于数据库的实现相同的交换算法。这两种方法产生统计上不可区分的网格。与原生实现相比，基于TSTT的交换组件的开销可以低至约50%;还给出了进一步减少此开销的建议。
Software re-use is an increasingly common practice in development of scientific computing software. Recently, there has been an emphasis on the definition of software components: software objects that use a clearly defined interface to encapsulate a specific functionality. In particular, at least two such component definitions have been developed for low-level mesh functionality and at least one of these already has several working implementations. These components open the door for the development of mesh manipulation services — such as swapping, smoothing, edge collapse, face splitting, etc — that are independent of the internals of the underlying mesh database; in turn, higher-level operations can be built on top of these. This paper describes a software component for simplicial mesh swapping in both two and three dimensions. Applications can control the cell quality measure used as a swapping criterion, including defining new quality measures. The swapping component is coded using the Terascale Simulation Tools and Technologies (TSTT) mesh component, and is therefore usable with any mesh database that implements the TSTT mesh interface. Results are given comparing the new software component with a database-specific implementation of the same swapping algorithm. The two approaches produce statistically indistinguishable meshes. The overhead for the TSTT-based swapping component can be as low as about 50% compared with the native implementation; suggestions for reducing this overhead even more are also given.