基于视频的流体表面重建与流固耦合仿真

发布时间：2018-04-04 13:07

本文选题：三维重建　切入点：流体仿真　出处：《华东师范大学》2016年硕士论文

【摘要】：气体,液体,多相流和混合流等流体现象,具有极其复杂的运动规律,而流体重建和仿真在工程应用与理论研究等方面必不可少。波涛汹涌的海浪、洪水决堤、泥石流灾难等流体仿真效果在电影游戏特效制作、灾难模拟防治、国防科技等领域中得到了广泛的应用。传统生成流体动画方法主要是基于物理的流体仿真方法,这类方法从某个初始状态开始,通过数值积分的方式来仿真流体力学,算法效率较低。另一方面,基于视频的流体重建方法可以较方便的重建流体表面,获得流体表面的几何信息。但是此类方法重建的表面只有单层数据,不能在物理模型驱动下与周围的环境进行双向交互。本文首先采用明暗恢复形状方法(SFS, Shape From Shading)对视频中的流体表面几何信息进行快速重建,获得每帧输入图像的表面高度场。然后将高度场与浅水方程结合,通过最小化能量方程的形式计算得出流体表面的速度场。重建的数据(包含高度场和速度场)作为真实数据来驱动基于物理的流体仿真。仿真阶段首先将表面几何信息作为边界约束条件离散化到整个三维体内部,获得体数据(多个带有位置和速度信息的体粒子)。然后将视频中重建的体数据导入到SPH仿真场景,作为仿真场景的初始条件,并与场景中的其他虚拟环境物体进行交互。为了解决物理仿真偏离重建数据的问题,不断使用下一帧重建的体数据来校正仿真场景数据。最终获得的流体交互动画中,非交互区域流体接近视频中重建的流体表面,而局部区域加入了新的交互效果,整个仿真场景由物理模型与重建的数据共同校正。本文对基于视频的流体表面重建与数据驱动的物理仿真的结合性进行了探索,充分利用流体视频中重建的信息,并结合SPH物理模型来进行流固耦合仿真。相比于传统重建与仿真方法,可以获得更接近于流体视频数据的流固耦合仿真效果。
[Abstract]:Fluid phenomena such as gas, liquid, multiphase flow and mixed flow have extremely complex motion rules, and fluid reconstruction and simulation are necessary in engineering application and theoretical research.The turbulent waves, flood, debris flow disaster and other fluid simulation effects have been widely used in the field of film game special effects production, disaster simulation prevention, defense science and technology and so on.The traditional method of generating fluid animation is mainly based on physical fluid simulation. This kind of method starts from a certain initial state and simulates fluid dynamics by numerical integration. The efficiency of the algorithm is low.On the other hand, the video-based fluid reconstruction method can easily reconstruct the fluid surface and obtain the geometric information of the fluid surface.However, the surface reconstructed by this method has only single layer data and can not interact bidirectional with the surrounding environment driven by physical model.In this paper, the shape restoration method, Shape From shading, is used to reconstruct the geometric information of fluid surface quickly, and the surface height field of each frame input image is obtained.Then the velocity field of the fluid surface is calculated by minimizing the energy equation by combining the height field with the shallow water equation.The reconstructed data (including height field and velocity field) are used as real data to drive physical fluid simulation.In the simulation stage, the geometric information of the surface is discretized to the whole three-dimensional body as a boundary constraint condition, and the volume data are obtained (many volume particles with position and velocity information) are obtained.Then the reconstructed volume data in the video is imported into the SPH simulation scene as the initial condition of the simulation scene and interacts with other virtual environment objects in the scene.In order to solve the problem of physical simulation deviating from reconstructed data, the volume data of the next frame is constantly used to correct the simulation scene data.In the final fluid interactive animation, the non-interactive region fluid is close to the reconstructed fluid surface in the video, while the local area is added with a new interactive effect. The whole simulation scene is corrected by the physical model and the reconstructed data.In this paper, the combination of video-based fluid surface reconstruction and data-driven physical simulation is explored to make full use of the reconstructed information in fluid video and SPH physical model to simulate fluid-solid coupling.Compared with the traditional reconstruction and simulation methods, the fluid-solid coupling simulation results are more similar to the fluid video data.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;O35

【相似文献】