Documentation Project| Terrain Technology |
 |
 |
 |
| Written by Administrator | ||||||||
| Tuesday, 12 August 2008 10:00 | ||||||||
| | The terrain height map rely on real World data, accumulated by NASA satellites (the so called Digital Elevation Maps, SRTM/GTOPO30 data sets), covered by high resolution satellite images of according areas. As you may imagine, the entire area includes many countries, even more air ports, still more cities and so on. While giving you a huge "play ground" (you will have unlimited freedom of movement on and over entire territory), SL Team provides real world features like cities, infrastructure etc. only for the area, which is in the scope of dynamic campaign. Even limited by campaign extent, SL represents 3 countries, 322+ air ports (more than in Falcon4), countless cities, villages and infrastructure.The populated areas and other features are of course of generic type, though alignment of the runways and coordinates for all ground features are correct. Experience shows, that communities often were willing to enhance multimedia content of their favorite simulations, this is why I leave software architecture as open as possible. The user will be able to add/modify facilities, textures and add their own 3d content to the game. Therefore I will provide a set of converters, animation sequencers and data base editors Click on images for greater resolution.
Update 2009-05-11: Thanks to continuous effort of the CSI-CGIAR organization, the SRTM30 data set receives year by year greater detail, better interpolation and smoother details. Out project moves to the 4.2 version of the SRTM30 data set (same digital elevation maps as you might have been pleased to notice in Google Earth). The raw set consists of ca. 852 zip archives containing ASCII height data with overall size of 14GB. A single ASCII file "weights" 250MB unpacked. It will be a demanding multi-core task to process those files. Our initial commitment will consist of a MySQL database to take unprocessed information for alter integration into our project. C-LOD and ROAM algorithms were considered for terrain visualization within SimuLab project.
Update 2009-06-25: It took us some iterations to find proper approach on using this tremendous data set. At initial steps we just tried to get the information contained in the archives into machine readable format. Extracting ZIP files and parsing text files for getting height information has never been considered seriously, so we looked for a simple method of getting ASCII data into mashine readable format. The first implemented import routine produced binary dumps of the values converted into numeric presentation. The files however were way too large for practical application: 140MB each the files sumed up to 120GB for entire world map. This incredible amount of data would never be distributable, so we looked for suitable compressing. There are many compressing algorithms and implementations around, but we looked for the most elegant solution of this problem. Finally, the internet and associated graphics formats inspired us for the proper solution: certain portions of the geo grids had same or insignificant variable height fields. Compressed format libraries like GIF, JPG, PNG have proper algorithms to detect such regions and describe them place efficient in the resulting image file. The second iteration now exported the height data into PNG images. Static gradient with distinct values for every height value was used to map height value to color. The resulting maps look very nice and have potential to be used for topographic and infrastructure editing needs later in the project. Still, some problems have been discovered in this approach: gradient generated by a CAD programm was pretty "unsharp" merging 3 to 5 pixels (in our terms 3 to 5 differnt height measures) to the same color. Generated map leaked exactness we would require to actually present the terrain in 3D.
Final import process sharpening went on to eliminate this problem. At the current stage, the importer produces color values programmatically to ensure loss free storing of the height data in the result image. Final output is an ugly looking tile containing exactly mappable color to height pixels still profiting from compression mechanisms associated with PNG file format. Finally, the process has been polished to partition full size grids containing 6000x6000 measured points into 1000x1000 pixel images and to run 4 conversions at the same time. Compression impact: we were able to reduce the data size from initial constant 250MB ASCII text to vairable 30MB to 4MB ranging PNG graphic file
|
|||||||
| Last Updated on Friday, 23 April 2010 12:22 |
