基于语义分析的三维模型生成技术研究

项目名称： 基于语义分析的三维模型生成技术研究

项目编号： No.61202333

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： 徐凯

作者单位： 中国人民解放军国防科学技术大学

项目金额： 25万元

中文摘要： 三维模型生成一直是计算机图形学领域的难点问题之一。本项目立足于人造物体三维模型所具有的丰富语义信息，围绕模型生成问题的两个重要方面，即面向低成本设备的三维模型快速获取与重建，以及面向创造式建模的智能启发与建议生成，研究基于语义分析的低质量扫描数据三维重建和三维模型集自动演化。本项目通过对人造物体三维模型集的联合语义分析，研究语义增强的结构化模型表示和面向语义的模型集一致分割，并以此为基础解决三维重建中的数据不完整性，以及三维模型集演化中的模型自动融合等关键技术难题。力争形成创新性和实用性兼具的三维模型生成方法，为三维建模领域做出基础性和前瞻性的贡献。

中文关键词： 三维模型生成；语义驱动的形状分析；数据驱动的几何处理；三维重建；三维扫描

英文摘要： 3D model creation is a long-standing problem in computer graphics. This project is interested in the creation of 3D models of man-made objects which contains rich semantic information. In particular, we will focus on the two important aspects of the model creation problem. The first aspect is the rapid acquisition and reconstruction of 3D models based on low-cost 3D scanning devices. The second one is the automatic generation of modeling suggestion and inspiration for creative 3D modeling. We propose to address these two challenges through the co-analysis of 3D model sets of man-made objects. Especially, we aim to propose novel semantics-augumented structural representations for 3D models and new methods for consistent segmentation of 3D model sets. Based on the semantic information pre-analyzed from the input set, we will try to address the key issue of data imperfection during the 3D reconstruction. At the same time, we propose to generate modeling inspiration through semantics-driven model set evolution, where the key challenge is how to blend two man-made shapes. This project aims at coming up with model creation methods that are both techincally novel and pratically useful, and strives to making substantial contribution to the state-of-the-art.

英文关键词： 3D model generation；Sementics-driven shape analysis；Data-driven goemetry processing；3D reconstruction；3D scanning