Particle System-based Real-time Fireworks Simulation Using GPU Acceleration

Fire,smoke,frog and fireworks are typical irregular fuzzy objects in natural world,which makes them difficult to be modeled by the conventional methods.Up to now,particle system is the most successful model for the simulation and generation of this type of fuzzy objects.Especially,fireworks has many types,vivid colors and various shapes.This makes the simulation of fireworks much more challenging.Particularly,with the enrich of fireworks types and the increase of fireworks exploding scope,even professional firework engineers lack an effective technical method to simulate the firework explotion following the topic and scene of fireworks.Thus,the existing firework explotion is usually designed by the experience and imagination of firework engieers,which can not guarantee the balance between final visual perception and cost.Meanwhile,because there are too many particles involved in the particle system based fireworks simulation,its computation is very intensive.And graphics processor unit(GPU)has been widely used in general-purpose computing applications such as graphics rendering.Following the practical requirements of large-scale firework simulation,this thesis investigates the particle system based fireworks simulation approach and its realization.Its motivation is to fully exploit the parallel computing capability of GPU,and to put the most computation-intensive operations such as refreshing the particle properties into GPU by the Compute Unified Device Architecture(CUDA)of Nvidia.Thus,a better balance can be achieved between the sense of reality and the realtime performance.The main works and contributions are summarized as follows.First,after analyzing the shapes of various types of fireworks and their classification,a univeral application programming interface(API)is designed for particle system,which is especially suitable for the large-scale fireworks simulation.Furthermore,the data structure of linked list is exploited for the large-scale fireworks simulation.Second,the most popular Li Hua Dan is chosen as the emphasis of firework simulation.A particle system-based real-time fireworks simulation approach is proposed by CUDA programming.It discusses in detail the initialization of particle system,the initialization of particles,the refreshing of particle properties and texture mapping.Moreover,the most computation-intensive operations such as the refreshing of particle properties are put into GPU platform to be accelerated by CUDA programming.With the aids of other techniques such as texture mapping and color blending,this thesis realize the fireworks simulation under the situations of single type,multiple types and multiple types with bitmap background.Experimental results show that it achieves desirable results and the rendering efficiency is also improved when it is compared with the simulation on the platform of CPU.Moreover,the more particles involved,the more rendering efficiency is improved.Therefore,this research work verifies the acceleration of fireworks rendering on GPU by exploiting CUDA programming.