Self-luminous Particle System Real Time Lighting Simulation

Particle system is one of the computer graphics techniques that simulate certain kinds of fuzzy phenomena which consist of a large number of tiny objects. And the tiny objects are moving and changing according to certain rules. Particle system is widely used in computer simulation. It can simulate many kinds of fuzzy natural phenomena.Many kinds of fuzzy phenomena simulated by particle system are self-luminous, such as firework, explosion and so on. The particle in these phenomena contains the property of self-luminous. The particle’s self-luminous property not only determine the brightness of the particle during computer rendering process, but also illuminate the surroundings, especially for firework and explosion which is indoor or in the dark scenario. Previous research only focus on particle system itself, the illumination by self-luminous particle system is ignored.This thesis takes a thorough research on the topic of the illumination by self-luminous particle system, including design and implementation of self-luminous particle system. After took a research on cluster partition algorithms and rendering methods, we set up a simulator framework which can simulate self-luminous phenomena. The main researches and contributions of this thesis include the following aspects:(1) Proposed a model to implement self-luminous particle system, focus on implementing the illumination brought by the self-luminous particles in the system to the surroundings and the flicker of the illumination. Each particle in the particle system was viewed as a point light. Then we built cluster partition model and use our cluster partition algorithm to decrease the lights number. Finally, we used tile-based deferred rendering method to render the scene with the big point lights. While tile-based deferred rendering method takes an appropriate process on the screen space information, it has a big improvement on rendering efficiency.(2) Presented a random sampling based cluster partition algorithm, in order to solve the computational efficiency problem brought by the large number of lights. Large numbers of point lights were divided into a small numbers of clusters, and then all of the point lights in a cluster were merged into a larger point light. We defined the dissimilarity degree between two point lights. Then we sampled a part of the point lights as seed lights based on the dissimilarity degree. At last, a specified number of clusters were generated by assigning all the other point lights to a seed light with lowest dissimilarity degree.(3) Brought out a division-based cluster partition algorithm. Random sampling based cluster partition algorithm can partition all of the lights in the scene into a specified number of clusters quickly. However, as it is based on random sampling, it may bring uneven sampling condition that some clusters are too big while some clusters are too small. This leads to huge cost of the total cluster merge. So we proposed a division-based cluster partition algorithm. It can effectively solve the problem brought by uneven sampling.(4) Based on self-luminous particle system, we implemented the effect of firework and explosion. After analyzed the experiments under different particles number and different sampled lights number, we found that our method can simulate the illumination and the flicker of illumination in real time. The experiments validated the real-time and effectiveness of our algorithms.