特征保持的三维建模及优化表示

Rapid and high-precision 3D modeling is a hot research topic in computer vision and graphics. It has so many important applications such as industrial design, digital city and national defense. However, the existing technologies have many disadvantages, for example, the reconstruction time is long, reconstructed point cloud data is redundant, the quality of the reconstructed model is low and maintaining geometric features is very difficult..According to the fact that feature curves are the simplest and most direct way to describe a geometric shape, this project studies 3D modeling techniques combining three-dimensional point clouds with feature curves. We aim at designing more reliable and more efficient 3D digital algorithms from a theoretical perspective. We focus on three key scientific issues, including rapid reconstruction of 3D curve network, surface reconstruction driven by 3D feature curves and optimization for the high-quality model. The research contents contain (1) rapid reconstruction of 3D feature curve network; (2) surface modeling based on 3D point cloud and feature curve network; (3) high-quality surface mesh optimization. This project innovatively embeds 3D feature curves as constraints into the surface modeling process. We are going to make a breakthrough in the basic theory and key technologies of high-quality modeling driven by 3D feature curves. Our project team has accumulated solid work, and this project can directly serve many intelligent applications, like rapid urban reconstruction, intelligent modeling on mobile terminals, virtual reality, etc.

快速和高精度的三维建模是计算机视觉和图形学交叉的热点研究话题，在工业设计、数字城市和国防军事等领域具有重要的应用价值。然而，现有的技术存在重建耗时过长、重建数据量大、重建模型质量差而且几何特征难以保持等缺点。.根据特征曲线是描述一个几何形体最简单和最直接的元素这一特点，本项目研究三维点云与特征曲线结合的三维建模技术，从理论角度设计更可靠、更高效的三维数字化算法，着重解决三维曲线网络的快速重建、三维特征曲线驱动的点云表面重构和高质量模型优化这三个关键科学问题。研究内容包括：1）三维特征曲线网络的快速重建；2）结合三维点云与特征曲线网络的表面建模；3）三维形体的优化表示。本项目创新性地将3D特征曲线作为约束嵌入到表面建模的过程中，拟在特征曲线驱动的三维高质量建模的基础理论和关键技术方面做出突破。项目组前期工作积累扎实，研究成果可服务于城市快速重建、移动终端智能建模、虚拟现实等智慧场景应用领域。

快速和高精度的三维建模是计算机视觉和图形学交叉的热点研究话题，在工业智能制造、智慧城市管理、遥感测绘和国防军事等领域具有重要的应用价值。本项目瞄准实际产业应用对高精度三维场景几何模型的迫切需求，研究保持几何特征的高效率和高质量三维建模技术。研究内容包括三维特征曲线网络的快速重建、结合三维点云与特征描述的几何建模、三维形体的优化表示。在执行过程中，本项目重点探索了二维图像特征线提取及三维曲线网络重建、三维建模及模型结构优化、精细树木几何建模、三维模型的特征计算等内容，取得了一些关键性的技术和成果。研究成果可期望直接服务于城市场景重建、动态仿真、地理信息、工业生产等众多领域。本项目在执行过程中共发表国际期刊论文16篇（均被SCI收录），国际会议论文6篇，国内核心期刊论文2篇，申请中国发明专利5项，已授权3项。在本项目的资助下，共培养博士研究生4名（均已毕业）、硕士研究生4名（均已毕业），博士后1名（已出站），职称晋升人才1名。项目组提出的核心关键技术已在华为、中国建筑设计研究院、深圳大学等企业获得成功应用，企业委托项目总经费近350万元。