面向大规模三维模型集的交互式共分割及应用

项目名称： 面向大规模三维模型集的交互式共分割及应用

项目编号： No.61202222

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 汪云海

作者单位： 中国科学院深圳先进技术研究院

项目金额： 25万元

中文摘要： 三维模型库的大量涌现和数据驱动方法的发展加大了人们对已有三维模型资源的利用程度，但多数基于模型库的应用都需要先从三维模型集（一组语义属于同一类但形状差异很大的三维模型）中提取模型的高层语义信息。作为提取这些信息的方法之一，共分割能同时对模型集中的所有模型进行分割并建立对应关系。但是目前的共分割方法多是自动分割，无法从复杂的模型集中提取准确的信息。因此，本项目拟研究交互式共分割，通过人机交互，将用户的先验知识融入到共分割中，使共分割更加准确高效。同时，本项目还将研究多用户协同交互式共分割，以解决单用户在面对大规模三维模型集时无法快速完成共分割的问题。设计面向三维模型集的交互式共分割算法和协同交互式共分割算法是本项目的关键问题所在，其目标是探索和发展面向大规模三维模型集的交互式共分割的关键理论和整套算法，开发新的基于模型集的应用，丰富人们对三维模型的理解和分析，从而提高模型库的利用率。

中文关键词： 共分割；模型集；语义分割；主动式共分析；投影分析

英文摘要： Due to the emergence of shape repositories and the development of data-driven methods, existing shapes are widely used in many computer graphics applications. However, before reusing shapes, most of applications based on shape repositoreis require us first to extract their high-level structures from a set of shapes which belong to the same class but exhibit significant shape variability. Co-segmentation is one of the methods that allows us to extract such information, since it reveals semantic shape parts and establishes their correspondence across the set. However, most of the existing methods work in an automatic way, which are not accurate enough on large sets of shapes.In this project, we study interactive co-segmentation, which incorporates the user knowledge into the co-segmentation so as to improve the accuracy of the co-segmentation result. We also aim at developing multi-user collaborative interactive co-segmentation, in order to address the difficulties in the co-segmentation of a large set of shapes with a single user. Thus, this project involves the design of the interactive co-segmentation flow and collaborative interactive co-segmentation algorithms. The major contribution of this research is to develop systematic theory and effective algorithms for the interactive co-segmentation of a large set of

英文关键词： co-segmentation；shape set；semantic segmentation；active co-analysis；projective analysis