基于草图的几何处理和应用

With the rapid development of 3D acquisition devices and advances in geometry processing algorithms, it is now easy to obtain highly complex 3D models or even 3D scenes, which makes it desirable to develop more advanced geometry analysis tools to adapt the fast growing rate of geometric data. Recent researches are focusing more and more on aspects regarding object structure and functionality. Automatically analyze and process such complex data is, however, very challenging to perform. In many cases, it is simply very difficult to obtain a semantically meaningful understanding of such geometric data. For example, learning the functionality of an object is known to be extremely difficult for computers. ..Such semantic information, however, is easy to acquire in our human vision system. In this proposal, we develop interactive sketching tools to assist such geometry processing tasks and subsequently enable intuitive and easy manipulation of complex geometric data. The key idea is that while people are hard to express their ideas through visual communication system, sketching, being inherently intuitive, fast, and easy-to-perform, can largely help ease the process when integrated with semantics. In particular, we are interested in the following two specific problems: interactive sketch interpretation for preparatory design studies and sketch-based geometry exploration and structure analysis for data collections. We design specific frameworks for both tasks, and look forward to linking the two tasks in a complementary framework. The success of the research will open up new directions for the development of interactive techniques in geometry analysis and exploration.

随着计算设备的发展，获取大量的高复杂度的三维数据变得越来越容易。基于此，处理这类数据的算法也随之日新月异。越来越多的研究开始着眼于对复杂物体以及大规模场景做结构和功能性方面的理解和分析。全自动的几何处理算法往往在这类复杂的数据集上面工作得不好或者不鲁棒。如何让计算机去自动理解和处理这些复杂的数据是目前几何处理一大难题，比如说，让计算机去自动识别一个物体的用处。..让计算机去自动理解功能结构性这类涉及语义学方面的信息的时候，一般从两方面入手，第一是通过大量的数据集训练，第二是用户人为的告诉计算机，就是我们所谓的先验知识。本项目的提出旨在通过用户草图勾勒这种简单的人机交互行为结合结构功能语义方面来帮助解决几何处理里面的一些重要问题。更具体的讲，本项目将包含两块内容：第一个是基于草图的三维场景结构分析，理解和建模以及基于内容的三维作业，第二个是基于草图和功能结构的三维模型数据库分析，浏览和组织，即通过用户简单的草图勾勒模型部件和标注部件之间相关的语义功能关系，来实现对大型三维模型库的实时浏览和基于部件的组织管理。

随着计算设备的发展和几何处理算法的日新月异，对于现今而言，获取大量的高复杂度的3D数据已经变得越来越容易了。正是如此，几何处理与几何理解算法的需求也越来越迫切。几何在计算机中的呈现方式多样，如何快速有效的让计算机能够去智能的识别和处理复杂几何信息是一大难题，比如让计算机去自动检测一个物体的物理功能。因此处理这类涉及语义学方面的信息的时候，往往通过一定的交互可以让建模结果变得更加鲁棒和符合人们意图。.本项目旨在利用人机交互的草图信息来更加有效的分析和处理几何信息。本项目的关键动机在于，设计语义建模的时候，上下文语义以及一些简单的草图输入往往能让分析变得容易且结果变得鲁棒。更具体的来讲，本项目拟提出基于用户草图输入的三维分析与建模的图形学应用。我们希望通过用户简单的勾勒草图的角度来试图尝试解决图形和视觉领域经典的难题：三维建模。