Construction of FortIrwinTerrain.wrl


CAPT Timothy D. McLean USMC and LCDR Marlon L. Smith USN

25 May 1999

The FortIrwinTerrain.wrl terrain database was created with Multi-Gen Creator software.  First, Digital Terrain Elevation Data (DTED) Level 2 data produced by the National Imagery and Mapping Agency (NIMA) was converted from its original format to Multi-Gen's proprietary Digital Elevation Data (DED) format. The Multi-Gen software required the conversion.  DTED Level 2 was chosen because it was the highest resolution unclassified elevation data available.

After the conversion, a gaming area was chosen.  The gaming area had to be large enough so the end of the gaming area could not be seen.  Also, sufficient mountainous terrain had to be available for interesting terrain flying.  For the above reasons, the Fort Irwin reservation was selected.  The latitudes and the longitudes of the gaming area follow.
 

Gaming Area Latitude-Longitude Coordinates
NW corner
NE corner
SW corner (origin)
SE corner
  35~30’00”N
117~00’00”W
   35~30’00”N
 116~30’00”W
  35~00’00”N
117~00’00”W
  35~00’00”N
116~30’00”W

These points correspond to an area 45512 X 55560 meters giving an area of approximately 2500 square Kilometers (Km2).

Next the number of polygons was chosen.  The real-time system that the model was to be rendered with was considered, as was the resolution of the terrain itself.  The first attempt was to build a database with 5000 polygons, however 5,000 polygons did not provide enough relief between elevation posts.  Ridges and valleys within mountainous terrain were not accurately portrayed.  The next model was built with 8,000 polygons and was barely acceptable for terrain recognition.  The last model was built with 50,000 polygons. A four-processor SGI Onyx Infinite Reality worked eight hours to complete the model.  No levels of detail were incorporated into the model at this point.  The model presented in this project is the 8000 polygon model because of the low-end PC's that need to render the model.  The polygon terrain skin itself is not acceptable for terrain recognition training.  However, draping the terrain skin with satellite imagery as a texture dramatically increases the terrain database's value for training.  The reason for this is that the shadows on the texture fool the eye into thinking valleys, ridges, and dips are actually in the model when the are not.

The texture image  FortIrwinTerrainTexture.jpg comes from Controlled Image Base (CIB) 5 meter imagery produced by NIMA.  In particular, the NIMAMUSE tool Raster Importer takes raw CIB data and produces a digital photograph in standard graphics format. Since the data was CIB5, every pixel represents 5 meters on the ground.  Measuring the area of the gaming rectangle, and dividing by 5 yields the correct number of pixels for an RGB formatted graphics image.  Unfortunately, it also produces a black and white texture over 100 megabytes.  Adding color to the texture for a more realistic environment produces a 300 megabyte image.  A jpeg was chosen as the output because the compression inherent in the jpeg format produced a texture small enough for download over a 10 Megabit network.

Since the texture was black and white, color was added so the user would be more immersed in the virtual environment.  Adobe PhotoShop 4.0 was used to add this color.  A brown hue was decided upon because it seemed the most realistic for the area and it was easy to implement using PhotoShop's colorize feature.  After several attempts a light brown hue was finally found that seemed to provide a nice balance between the darker mountains and the whiter dry lake beds.

Also, the CIB image inherited two darker strips of overexposed areas where the satellite overlapped on successive data collection runs.  These strips had to be removed because they were distracting to the user during higher bird's eye views.  PhotoShop's rubber stamp tool was used to remove the dark strips leaving a barely noticeable strip of repeated texture.

The texture was applied to the terrain skin using Multi-Gen's 4 point put method.  This method matches the four corners of the texture with the four corners of the terrain skin and automatically applies the texture to each polygon of the terrain skin.

The completed terrain database was still not in a format recognizable by VRML renderers so the last step is to use Multi-Gen's export feature that converts Multi-Gen's Open Flight format to VRML 2.0.  With the completion of the last step, the VRML terrain model was produced and handed over to Marlon Smith for further incorporation into the project.

5 Meter Controlled Image Base (CIB5) is an unclassified/limited distribution seamless dataset of orthophotos, made from rectified grayscale aerial images collected from national sensors and degraded, resulting in a ground sample distance (resolution) of five meters. CIB data is produced from digital source images and are compressed and reformatted to conform to the Raster Product Format (RPF) standard. CIB files are physically formatted within the National Imagery Transmission Format (NITF) message. Applications for CIB include rapid overview of areas of operations, map substitutes for emergencies and crises, metric foundation for anchoring other data in C4I systems or image exploitation, positionally correct images for draping in terrain visualization, and image backgrounds for mission planning and rehearsal.

Digital Terrain Elevation Data (DTED) is a uniform matrix of terrain elevation values. It provides basic quantitative data for all military systems that require terrain elevation, slope and gross surface roughness information. Data density depends on the level produced. DTED0 post spacing is 30 arc seconds (approximately 1000 meters). This corresponds to a small scale hardcopy product. DTED1 post spacing is three arc seconds (approximately 100 meters). This corresponds to a medium scale hardcopy product. DTED2 post spacing is one arc second (approximately 30 meters), corresponding to large scale hardcopy products.

Collision detection and terrain following over this geometry is performed by CollisionDetectionTerrainReader.java which was developed by Marlon Smith as part of the CS4472 Advanced Physically Based Modeling course.

Reference:
McLean, Timothy D., An Interactive Virtual Environment for Training Map-Reading Skill in Helicopter Pilots, Master's Thesis, Naval Postgraduate School, Monterey California, September 1999.  Available via www.npsnet.nps.navy.mil/npsnet

This page is available at
http://www.web3D.org/WorkingGroups/vrtp/demo/helicopter/FortIrwinTerrainMetadata.html

Point of contact: Don Brutzman (brutzman@nps.navy.mil)