Evolutionary design of two-dimensional fractals and three-dimensional plant structures for computer graphics.

摘要

This thesis deals with the evolutionary design of 2D fractals and 3D plant structures using genetic programming. The ultimate goal of this research is to enable the user, who is not required to understand the underlying model creation processes, to generate 3D plant models.;An interactive evolutionary design system, Breeder, is developed for the evolutionary generation of graphics models. A population of rule-based graphics models is generated or modified in Evolvica and visualized in cpfg. Evolvica is a Mathematica based evolutionary algorithms environment. cpfg is a rule-based plant modeling environment. An evaluation panel allows users to interactively evaluate models. The whole visual design process can be exported into a Mathematica notebook document.;Evolution of 2D fractal curves using Breeder is carried out with the motivation to illustrate the principle of the evolution of L-systems and therefore serves as an introduction to the more complex evolution of 3D plants.;For the evolutionary design of 3D plant models using Breeder , a general 3D plant representation is proposed to deal with both topology and geometry of plant structures. Furthermore, a template is discussed for evolving 3D plant models that conform to the general representation. As a result, various visually appealing 3D plant models, which are compatible with cpfg, are evolved.;The two applications have shown that Breeder provides a general mechanism for the creation of rule-based graphics models.