In this thesis I examine the problem of modeling and visualizing large groups of plants. The extreme visual complexity of such scenes can be captured using multilevel models, which combine a hierarchy of models at different levels of abstraction. The modeling of plant communities is done by two-level models, where the higher-level model describes the spatial distribution of plants, and the lower-level model describes individual plants' shapes.
Models of plant communities can be divided based on the direction of information flow: local-to-global models are rooted in individual-based ecosystem simulations, while the properties of individuals in global-to-local models are inferred from a given distribution of plant densities. Multiset L-systems are introduced as a formalism for local-to-global models of the spatial distribution of plants. A new global-to-local model is developed which uses the idea of iterative deformation of a probability density.
Examples of both local-to-global and global-to-local models of the spatial distribution of plants are demonstrated which exhibit various ecological properties, including succession, clustering, and self-thinning. These examples are coupled to a low-level model of individual plants to create realistic visualizations of the plant community.
Brendan Lane. Models of Plant Communities for Image Synthesis. M.Sc. thesis, University of Calgary, June 2002.
Download PDF here (15 Mb)Back to Publications