One of the fundamental modules in D5 supplies a model of the earth (or other, roughly ellipsoid planet) to the core representation of the DIRSIG universe. The distributed version of D5 will likely contain a simple, material-mapped sphere representation of the earth by default (to keep download sizes reasonable), with the option to obtain more detailed (and more correct) models.
To this end we've been working on an earth model based on the WGS84 reference ellipsoid and SRTM (Shuttle Radar Topography Mission) DEM data. Since this is a very large data set (the 500m data we're using currently is 4.7GB in its raw form) it is an ideal case to test some of our new large file handling routines and integrating them into geometry interactions.
We model the earth as a collection of triangle mesh patches representing cells (primarily hexagons) of a geodesic, specifically the ISEA aperture 4 resolution 7 grid (i.e. 163,842 index cells). Each patch can be loaded into memory independently, pre-optimized for rendering, giving us a way to handle the large data set smoothly. The resulting distribution of patches looks like:
The SRTM data is then sorted into index cells. The plot below shows a small subset of the DEM near the northwest shore of Lake Erie sorted into index cells:
Once sorted, we then generate a mesh within the index cell with approximately one triangle per DEM posting. The triangles are equal area, and with the exception of the 12 pentagons in the geodesic, each cell has the same number of triangles (this will facilitate further optimizations for GPU rendering). A single cell mesh will look something like this (lines represent facets, the points are the original SRTM data points):
We are currently in the process of formalizing this pipeline and incorporating the final result into the core model. Note that we don't expect users to want to routinely use the earth DEM as the basis for traditional scenes (there is not enough detail here for most purposes). Scenes can either be built on the earth model terrain or replace it entirely with their own. The earth DEM is primarily there to provide baseline geometry for background radiance (including important shadowing effects) and to provide data for large pixel footprints.
To this end we've been working on an earth model based on the WGS84 reference ellipsoid and SRTM (Shuttle Radar Topography Mission) DEM data. Since this is a very large data set (the 500m data we're using currently is 4.7GB in its raw form) it is an ideal case to test some of our new large file handling routines and integrating them into geometry interactions.
We model the earth as a collection of triangle mesh patches representing cells (primarily hexagons) of a geodesic, specifically the ISEA aperture 4 resolution 7 grid (i.e. 163,842 index cells). Each patch can be loaded into memory independently, pre-optimized for rendering, giving us a way to handle the large data set smoothly. The resulting distribution of patches looks like:
The SRTM data is then sorted into index cells. The plot below shows a small subset of the DEM near the northwest shore of Lake Erie sorted into index cells:
Once sorted, we then generate a mesh within the index cell with approximately one triangle per DEM posting. The triangles are equal area, and with the exception of the 12 pentagons in the geodesic, each cell has the same number of triangles (this will facilitate further optimizations for GPU rendering). A single cell mesh will look something like this (lines represent facets, the points are the original SRTM data points):
We are currently in the process of formalizing this pipeline and incorporating the final result into the core model. Note that we don't expect users to want to routinely use the earth DEM as the basis for traditional scenes (there is not enough detail here for most purposes). Scenes can either be built on the earth model terrain or replace it entirely with their own. The earth DEM is primarily there to provide baseline geometry for background radiance (including important shadowing effects) and to provide data for large pixel footprints.
Comments