An integrated framework for physics-based simulation in computer vision

我们的研究最初集中在从距离(3D)数据建立人造物体的几何建模的基本问题上，用于离线建模。然后，我们解决了交互建模的问题。矢量场的概念被提出，并取得了科学和商业上的成就。这些年来，现有计算机的能力急剧增加，计算机图形学、虚拟现实、分布式计算和仿真等领域取得了重大进展。我们现在正在见证这些领域与计算机视觉领域的融合。如今，工作团队很少在不进行模拟的情况下设计复杂系统或在复杂环境中计划运营。为几何模型添加动态行为，使这些模型在逼真的模拟中交互，并为用户提供灵活的工具来探索模拟，这些都是具有挑战性的主题，激励了本提案中提出的计划，其长期目标是开发一个支持复杂3D几何环境中基于物理的模拟的集成框架。该计划由两个组件组成：i)旨在构建超大型结构(建筑物、城镇等)的几何模型的视觉组件，以及ii)旨在开发一种方法来向这些几何模型添加行为并在沉浸式虚拟现实环境中探索这些行为的虚拟现实组件。这一计划扩展了我们目前的研究范围，解决了领域之间的融合，并开辟了有希望的研究轨道。为了达到这一长期目标，研究将首先集中在四个短期目标：1-在视觉组件中，将矢量场框架扩展到超大型环境的几何建模问题；2-在虚拟现实组件中，开发沉浸式环境中仿真的交互式可视探索框架的基础；3-培训高素质人才；4-发布研究成果并将技术转移到用户部门。

Our research has initially focused on fundamental problems in geometric modeling of man-made objects from range (3D) data for off-line model building. We then addressed the problem of interactive modeling. The concept of vector field was proposed and has led to both scientific and commercial achievements. Over these years, the power of off-the-shelf computers has increased drastically and fields such as computer graphics, virtual reality, distributed computing, and simulation have made significant progress. We are now witnessing a convergence of these fields with the field of computer vision. Today, workteams rarely design complex systems or plan operations in complex environments without simulating them. Adding dynamic behaviour to the geometric models, making these models interact in realistic simulations, and providing users with flexible tools for exploring the simulations are challenging topics that motivate the program presented in this proposal whose long term objective is to develop an integrated framework for supporting physics-based simulation in complex 3D geometric environments.The program is comprised of two components: i) a vision component which aims at building geometric models of very large structures (buildings, towns, etc.), and ii) a virtual reality component which aims at developing an approach for adding behaviours to these geometric models and for exploring these behaviours in immersive virtual reality environments. This program expands our current line of research, addresses the convergence between fields, and opens promising tracks of investigation. To reach this long term objective, the research will first focus on four short term objectives: 1- in the vision component, expand the vector field framework to the problem of geometric modeling of very large environments; 2- in the virtual reality component, develop the foundations of a framework for interactive visual exploration of simulations in immersive environments; 3-train highly qualified personnel; 4- publish the research results and transfer the technology to the user sector.