Realistic tree modelling

Abstract

The main idea of the realistic tree modelling is to create the algorithm, utilizing the ground-truth data of a laser scanning during the virtual modelling. The reasonable approach deals with voxelization that evaluates a density of points in order to represent 3D grid cells as the high-density, middle-density, and zero-density structures. This permits to make the hypotheses about the individual tree shapes and a tree distribution in the forest. The promising techniques, such as space colonization algorithm, graph-based modelling, self-organizing tree modelling, and inverse procedural modelling, are discussed in details. The shape, geometric, and structural measures help to estimate the obtained modelling results. The experiments show that a shape distance between a modelling tree and a sample tree influences on the similarity significantly. The best results were received using the space colonization algorithm and inverse procedural modelling.