实时雪场景模拟技术的研究与实现

发布时间：2018-06-02 04:02

本文选题：雪模拟 + 粒子系统　；参考：《电子科技大学》2012年硕士论文

【摘要】：利用计算机图形学原理来真实的模拟各种自然现象已经是当今从事于虚拟现实仿真的研究者们的一大愿望。随着图形硬件的快速发展，越来越多的仿真系统应用于计算机游戏、电视广告和电影拍摄之中。雪景作为北国冬天必不可少的一种自然景观，没有雪景的实时模拟是无法反映出“千里冰封，万里雪飘”的磅礴气势的。目前，，在降雪场景模拟方面，更多的研究侧重于反映其真实感，而很少考虑实时性的要求；在积雪场景模拟方面，大多数的研究都倾向于渲染雪花沉积的过程，即从没有积雪到覆盖积雪这个过程，而在大多数实现中也往往很少考虑实时性的要求。本文将通过介绍并分析雪模拟的一些相关方法，对其原理作进一步探究，主要包括：粒子系统，雪花物理运动规律，降雪的纹理绘制以及基于深度图的静态积雪模型。通过研究并融合这些系统理论，使我们的雪场景模拟系统能够满足实时性的要求。本文的主要工作如下： 1）分析并讨论了降雪粒子在空中运动时的物理运动模型，并利用GPU相关技术，使得在每一次渲染更新前，能通过前一次降雪粒子的位置、速度等属性，根据雪花的物理运动规律实时计算降雪粒子的当前帧的相关属性，从而实现了在视点近景范围内的降雪模拟。 2）将动态纹理的相关理论应用于远景处降雪场景的实时绘制，提出了基于视点的双半椎体的绘制方法，并讨论了视差效果的实现手段。 3）提出了快速真实的静态积雪渲染方法。基于切比雪夫不等式的概率学理论，将不等式的概率上限近似作是对应点的积雪厚度比率，通过硬件滤波，模拟出比较具有真实感的“浅雪区”效果。 4）结合可编程流水线、LOD、点精灵、RTT等相关技术，详细地阐述了上述提出的相关算法的设计方案和实现方法，最后进行实践验证，实现了基于GPU加速的实时雪景模拟系统，实验结果表明，所实现的算法能够满足一定的实时效率的要求。
[Abstract]:Using the principle of computer graphics to simulate all kinds of natural phenomena is now a great desire of researchers engaged in virtual reality simulation. With the rapid development of graphics hardware, more and more simulation systems are used in computer games, TV advertising and film shooting. Snowscape is an indispensable natural landscape in North China in winter. Without the real-time simulation of snowscape, it is impossible to reflect the momentum of "thousands of miles of ice, thousands of miles of snow floating". At present, in the snow scene simulation, more researches focus on reflecting its reality, but few consider the requirement of real-time. In snow scene simulation, most of the research tend to render the snowflake deposition process. That is, from no snow cover to snow cover, but in most implementations, the requirement of real time is seldom considered. This paper will introduce and analyze some related methods of snow simulation, and further explore its principle, including particle system, snowflake physical motion law, snow texture rendering and static snow cover model based on depth map. By studying and merging these system theories, our snow scene simulation system can meet the requirement of real-time. The main work of this paper is as follows: 1) the physical motion model of snowfall particles in the air is analyzed and discussed. By using the GPU correlation technology, the position, velocity and other attributes of the previous snowfall particle can be passed before each render update. According to the physical motion law of snowflakes, the related properties of the current frame of snowfall particles are calculated in real time, and the snowfall simulation in the close-range of view is realized. 2) the theory of dynamic texture is applied to the real-time rendering of snowfall scene in the foreground, and a method of rendering double hemispheric body based on viewpoint is proposed, and the method to realize the parallax effect is discussed. 3) A fast and real static snow rendering method is proposed. Based on the probabilistic theory of Chebyshev's inequality, the probability upper limit of the inequality is approximated as the ratio of snow thickness at the corresponding point. Through hardware filtering, the effect of "shallow snow zone" with realistic sense is simulated. 4) combined with the relevant technologies such as programmable pipeline and point wizard, the design scheme and implementation method of the above algorithms are described in detail. Finally, the practical verification is carried out, and the real-time snow scene simulation system based on GPU acceleration is realized. Experimental results show that the proposed algorithm can meet the requirements of real-time efficiency.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41

【参考文献】