Procedural feature generation for volumetric terrains using voxel grammars

Abstract

A B S T R A C TTerrain generation is a fundamental requirement of many computer graphics simulations, including computer games, flight simulators and environments in feature films. There has been a considerable amount of research in this domain, which ranges between fully automated and semi-automated methods. Voxel representations of 3D terrains can create rich features that are not found in other forms of terrain generation techniques, such as caves and overhangs. In this article, we introduce a semi-automated method of generating features for volumetric terrains using a rule-based procedural generation system. Features are generated by selecting subsets of a voxel grid as input symbols to a grammar, composed of user-created operators. This results in overhangs and caves generated from a set of simple rules. The feature generation runs on the CPU and the GPU is utilised to extract a robust mesh from the volumetric dataset.

Flexible Terrain Erosion - A Fluid Simulation-Independent Approach Compatible with Multiple Representations

“…Our formulation for height fields (9), can be used to erode 2D scalar field-based representations. Similarly, our proposition for SDF (11) enables erosion for continuous 3D scalar fields and voxels (13) for discrete 3D scalar fields respectively.…”

“…Voxel Grids are regular, uniform volumetric grids that encode information on the presence or absence of material for each 3D point in the domain. Voxel grids are advantageous due to their regularity [13] and ability to represent volumetric models at the cost of high memory footprint, which has limited their use in terrain generation [21,23,26]. We consider two voxel grid representations : density-voxel grids for which each voxel contains a scalar value, for instance the occupation percentage [14] and binary voxel grids that can be seen as a mask containing the presence of material information.…”

Evaluation Metrics for Intelligent Generation of Graphical Game Assets: A Systematic Survey-Based Framework

“…The highlevel 3D categorisations include scenery, hard-surface and characters/creatures. Scenery groups together clouds [24], roads [25], terrain [26], trees [27] and rocks [28]. Hardsurface groups buildings [2], furniture [29], vehicles [5] and props [30].…”

A Multilevel Terrain Rendering Method Based on Dynamic Stitching Strips

“…There are various representations of terrain, such as voxels, procedural terrain, and height maps [6]. Considered as the 3D version of pixels, voxels are able to represent volumetric terrain features, such as caves and overhangs [11][12][13]. However, to represent a large and detailed terrain model, billions of voxels may be needed, which will cause intensive memory requirement and are inappropriate for real time rendering [14].…”