Research on modeling trees and plants has attracted a great deal of attention in recent years. Early procedural tree modeling can be divided into four main categories: rule-based algorithms, repetitive patterns, cellular automata, and particle systems. These methods offer a very high level of realism; however, creating millions of varied tree datasets manually is not logistically possible, even for professional 3D modeling artists. Trees created using these previous methods are typically static and the controllability of these procedural tree models is low. Deep generative models are capable of generating any type of shape automatically, making it possible to create 3D models at large scale. In this paper, we introduce a novel deep generative model that generates 3D (botanical) tree models, which are not only edible, but also have diverse shapes. Our proposed network, denoted BranchNet, trains the tree branch structures on a hierarchical Variational Autoencoder (VAE) that learns new generative model structures. By directly encoding shapes into a hierarchy graph, BranchNet can generate diverse, novel, and realistic tree structures. To assist the creation of tree models, we create a domain-specific language with a GUI for modeling 3D shape structures, in which the continuous parameters can be manually edited in order to produce new tree shapes. The trees are interpretable and the GUI can be edited to capture the subset of shape variability.

Shifting perspectives on restoration and management of public lands in the inland West have resulted in an increased need for maps of potential natural vegetation which cover large areas at sufficient scale to delineate individual stands . In this study, classification tree modeling was used to predictively model and map the plant association types of a relatively undisturbed wilderness area in the Blue Mountains of northeastern Oregon. Models were developed using field data and data derived from a geographic information system database. Elevation, slope, aspect, annual precipitation, solar radiation, soil type, and topographic position were important predictor variables. The model predicted plant association types with a relatively high degree of accuracy for most plant association types, with the lowest accuracy for the types within the grand fir series. Fuzzy confusion analysis was used to analyze model performance, and indicated the overall model accuracy was 72%.

This study integrated remotely sensed data, geographic information system (GIS), and classification tree-based modeling to delineate ecological units for the Ashley National Forest. Data points , provided by the Ashley National Forest, with a known location and dominant vegetation type, were related to data layers that were determined to be helpful in a landtype classification. These layers included elevation, slope, aspect, potential solar irradiation, precipitation, geology, basins, Landsat thematic mapper (TM) bands 3, 4, 5, and 6, and basic land cover. These points, with their related information, were then used to train the tree-based model for landtype classification. This resulted in a set of rules, in the form of a binary decision tree, that could be applied to the entire study area. After the landtype classification was obtained, it was cross-classified with geology to produce a landtype association layer. This resulting data layer was compared to an existing landtype association map and it was determined, by cross-tabulation, that the two classifications identified many of the same patterns.

The modeling of trees is challenging due to their complex branching structures. Three different ways to generate trees are using real world data for reconstruction, interactive modeling methods and modeling with procedural or rule-based systems. Procedural content generation is the idea of using algorithms to automate content creation processes, and it is useful in plant modeling since it can generate a wide variety of plants that can adapt and react to the environment and changing conditions. This thesis focuses on and extends a procedural tree generation technique that uses a space colonization algorithm to model the tree branches' competition for space, and shifts the previous works' focus from realism to fantasy. The technique satisfied the idea of using interaction between the tree's internal and external factors to determine its final shape, by letting the designer control the where and the how of the tree's growth process. The implementation resulted in a tree generation application where the user's imagination decides the limit of what can be produced, and if that limit is reached can the application be used to randomly generate a wide variety of trees and tree-like structures. A motivation for many researchers in the procedural content generation area is how it can be used to augment human imagination. The result of this thesis can be used for that, by stepping away from the restrictions of realism, and with ease let the user generate widely diverse trees, that are not necessarily realistic but, in most cases, adapts to the idea of a tree.

One of the most important challenges in real-time simulation of large trees and vegetation is the vast number of calculations required to simulate the interactions between all the branches in the tree when external forces are applied to it. This paper will propose the use of algorithms employed by applications like cloth and soft body simulations, where objects can be represented by a finite system of particles connected via spring-like constraints, for the structural representation and manipulation of trees in real-time. We will then derive and show the use of Verlet integration and the constraint configuration used for simulating trees while constructing the necessary data structures that encapsulate the procedural creation of these objects. Furthermore, we will utilize this system to simulate branch breakage due to accumulated external and internal pressure.

No description available.