Level Of Detail Trees

A level of detail tree allows you to implement multi-resolution meshes and switch between these meshes as the camera position changes.

Figure 1.1.Consider the following scene graph structure. This is an appropriate structure for two terrain quadrants.

The centers must be the same for the entire level-of-detail node in each tree or there will be gaps or overlaps when switching occurs. Since simplified geometry might have a different bounding box size than unsimplified geometry, you have to be careful when you're building level-of-detail trees by hand. [Scenome's terrain building tools automatically configure level-of-detail settings.]

Level-Of-Detail Center

The level-of-detail center is usually the centroid of the mesh. However there are cases where a different center location, such as a corner of the bounding box, might be more appropriate.

Figure 1.2. Level-of-detail center at the centroid. In this case the centroid is at 0,0,16. This graphic displays the level-of-detail center at the centroid.

The High Resolution Mesh

Figure 1.3. This mesh is displayed while the camera is between 0 and 512 feet from the center of the node.

This graphic displays an image of the high detail mesh.

The Medium Resolution Mesh

Figure 1.4. This mesh is displayed while the camera is between 512 and 1024 feet from the center of the node.

This graphic displays an image of the medium detail mesh.

The Low Resolution Mesh

Figure 1.5. This mesh is displayed while the camera is between 1024 and 100,000 feet from the center of the node.

This graphic displays an image of the low detail mesh.

Spherical Area Of Effect

Since level of detail culling uses a center value and a switch in/out distance, you can express its area of effect with a sphere. This sphere should encompass a lot of the mesh but does not need to fully enclose in most cases. However it's generally considered best practice to use a sphere that completely encloses the mesh.

Figure 1.6. The example below shows a mesh with a very close switch out distance. In fact, you can see that the mesh switches out while the viewer is still very close. The camera must be inside that sphere for the mesh to switch in. This value is nonsensical because the mesh won't be visible until the camera is well inside the mesh' bounding box.

This graphic displays an image of a 4x4 polygon rectangle sheet.

Figure 1.7. Using a larger switch out distance. The camera must be inside that sphere for the mesh to switch in. This value is appropriate.

This graphic displays an image of a 4x4 polygon rectangle sheet.