Studies On Realistic Modeling And Rendering Of Atmospheric Phenomena

Modeling and rendering of natural phenomena has been a hotspot and one of the mostdifficult tasks in Compute Graphics. It has been found wide application in manydomains such as special effects of movie, computer animation, battlefield simulationand virtual reality, etc. As a special class of natural phenomena, atmosphericphenomena are closely related to our daily lives, which are often included in movies,games and systems of view simulation, et al. However, it has been a challenge task tomodel and render atmospheric phenomena realistically due to their complex physicalmechanisms.This paper first introduces the research significance, research difficulties as well asbasic theories and methods of realistic simulation of atmospheric phenomena. Thensurveys related works about the simulation of atmospheric phenomena, and analysestheir advantages and applied limitations. In this paper, we focus on realistic modelingand rendering of several special atmospheric phenomena including realistic modelingand rendering of Buddha Glory, physically based simulation and animation of tornado,fast modeling and rendering of dynamic cloud scene, real-time generation of infraredsky scene.●Buddha Glory is a spectacle of nature, which has an important effect on Buddhistculture and tourism in Chinese history. In this paper, a novel method of physicallybased simulation of Buddha Glory is presented. We first propose a back scatteringmodel of water drops based on Mie scattering theory. Then, we calculate thespectral scatter intensity of glory rings by this model. Considering the solarelevation, the diameters as well as distribution of water drops in fog or cloud, wecalculate the shape and color of glory rings accurately. Next, according to thecharacteristic of shadow in Glory, we present a new shadow model to determinethe fuzzy and deformed shape of shadow in Glory. After converting the calculatedspectrum to RGB colors, we adopt the method of path scattering integral togenerate the whole attenuated scene of Glory. Finally, we have rendered differentGlory scenes under different atmospheric conditions at different times. Comparedwith the photographs of real Glory displays, our synthetic results are quitesatisfactory. ●Considering realistic light effects and tornado's damage to surrounding objects,we propose a novel physically based method of realistic simulation of tornadoscene. First, a Two-Fluid model is presented for modeling the flow of tornado. Asthe calculation of Two-Fluid system is more complex than single fluid system, wedesign a special solver for Two-Fluid model to achieve fast simulation. Taking thescattering and absorption of light by the participating media into account, theillumination effects of the tornado scene can be generated realistically. This paperhas also proposed a voxel connected model and successfully simulated tomado'sinteraction with large objects such as a car, a house, etc.●Propose a novel method of simulating dynamic cloud scene. We first model themotion of cloud based on the physical theory of cloud formation. Then, abovemodel of cloud is solved in discrete sparse grids. Thus we can get approximatedmotion of cloud. Next, we advect texture image by the calculated velocity field toadd the local detail of dynamic cloud scene. Here the texture image we used isPerlin noise map. The color of arbitrary position is got by calculating the productof the density texture and the regenerated Perlin noise texture. Finally, by usingdifferent texture, realistic dynamic cloud scene at different light environment canbe generated at high rendering rates.●Based on the theory of infrared radiation, we propose a method for real-timegeneration of infrared sky background. We first model the shape of cloud by themethod of particle system. Then, a radiation model of infrared cloud is presentedbased on the theory of atmospheric optics, infrared physics, etc. Finally, byadopting the hardware accelerating techniques, different infrared dynamic skyscenes could be successfully drawn in real time.