ACCURACY OF 3D (THREE-DIMENSIONAL) TERRAIN MODELS IN SIMULATIONS

Abstract

The usage of realistic three-dimensional (3D) polygon terrain models with multiple levels of detail (LOD) is becoming widespread in popular applications like computer games or simulations, as it offers many advantages. These models, which represent an actual location in the world, are essential for the simulation-based training of military vehicles like planes, helicopters or tanks. Because training scenarios on this kind of simulations are used to observe or to hit a target on the modeled location. In addition to that, driving the behavior of terrestrial vehicles is influenced by the terrain properties like slopes, ramps, hitches, etc. because of the direct interaction with the ground. For this reason, the terrain models in the simulation scene should not only display the textures realistically, but also represent an accurate morphology; meaning the terrain altitudes should be modeled as correct as possible. Such terrain representations can be created by using Digital Terrain Model (DTM) for the geometry and satellite images for texturing. The geometry models are in the form of polygonal meshes through the triangulation methods. However, the accuracy is influenced by some parameters. Using insufficient (under-refined) triangles during the 3D modeling causes missing of some altitude vertices. That means these points will not be present in the model. Consequently, it can be thought that the number of triangles should be increased for a better geometrical fidelity. Nevertheless, it is not always correct as the usage of too much (over-refined) triangles can also cause errors, especially in terrains with almost vertical faces (like cliffs). In addition to that, the performance of the system deteriorates drastically through the increase in the number of triangles, as the computational complexity is also getting higher.