PhD Thesis Defence (Aleš Zamuda)
Title: Diferencialna evolucija za rekonstrukcijo parametriziranih proceduralnih drevesnih modelov
Candidate: mag. Aleš Zamuda
Mentor: full prof. dr. Janez Brest
Date: 15 May 2012 ot 11:00
ZAMUDA, Aleš. Differential Evolution for Parameterized Procedural Woody Plant Models Reconstruction: PhD thesis. Maribor: A. Zamuda, 2012. XXII, 164 pg., ilustr. http://dkum.uni-mb.si/Dokument.php?id=30084. COBISS.SI-ID 16001302
In this dissertation, we convey thesis on our own development of a new evolutionary algorithm, which enables modelling of woody plants using parameterized procedural models reconstruction from image projections of reference models. We show, that after encoding the proposed procedural odel in a genotype, it is possible to create an evolutionary process to search for parameterized procedural models. By composing a fitness function and enotype-phenotype mapping, we show that differential evolution (DE) algorithm is suitable for iterative reconstruction of the chosen parameterizable rocedural model. We also show that the DE algorithm is especially useful for evolution of the chosen model, since it is of fixed dimension and encodes the arameters in real number domain, for which DE algorithm is efficiently designed. We also confirm, that jDE algorithm with self-adaptive control arameters is better than basic DE algorithm without self-adapting control parameters. Obtained results confirm, that the proposed approach is suitable or modeling of woody plants for computer animation, using evolution of a numerically coded procedural model. By this we confirm, that the thesis of this issertation confirms the set hypotheses. We have also published our findings displayed in this dissertation to a wider scientific public.