From: Curtis L. O. <cu...@fl...> - 2001-05-05 03:57:31
|
Martin Dressler writes: > So when CVS is not working, and I am out of compile cycle :o) I have some > time to simple sat before comp and think about things and I get some ideas. > > 1) Levels of detail > You can make tile with lower number of triangles and edge with tile at higher > LOD doesn't need to fit exactly, only is important that border edge of tile > on lower LOD is every time below the edge of tile at higher LOD. You will > never look under face because you will never be above tile with lower LOD. > You can also generate one texture for tile in some similar way, how it is > done in Atlas. For tiles that are far then ~10 km you don't need to texture > them and only paint them gray because eye see far object as grey ( it depend > on light conditions and size of object). > Buildings could be rendered from far distance as bilboard texture or as grey > bounding box. Steve Baker had a classic post on about a dozen different schemes for doing terrain LOD and their various strengths and weaknesses. This was pre-ROAM and friends. All of the techniques had been tried at various times and all of them left much to be desired. The last couple times I tried to locate that post I couldn't find it so perhaps it is lost forever. I have no problem with people experimenting with various LOD schemes, but do be aware that this stuff is very hard, and I think you will find it harder than you think to ensure that the entire edge of the further tile is lower than the nearer tile, especially considering that we can view these things from any angle. > 2) weather > How it look with weather from metar? Is it working. I have some idea how it > could be done. You should prepare two tables in metakit or other database > file, and load them on startup. In first table is list of all known metar > stations with record of last known weather and actual weather. In second > table are all triangles between neighbours metar stations (sorted in some > clever way to allow quickly find intersection). You then walk through list of > metar station and update actual weather and make this steps. (It should > run in new thread) > 1. get data > 2. data are too old? give station to unactive list > 3. is there new station? divide triangle in which station is, to 3 new > triangles, and delete old. > 4. get unactive list and for every unactive station find all neighbour > station (simply done by finding all triangles with unactive station) > and try to interpolate weather and save it in database. > > I hope that you find my ideas interesting. My english is terrible and if you > didn't understand to something I try to ask someone for better transaltion. METAR based weather is something I'd love to see developed. Christian's weather database has some fancy interpolating code, but it takes 'forever' to initialize at startup ... on the order of minutes... which is a lot. I'd be in favor of a less fancy scheme that was less CPU intensive ... perhaps pick the closest station and migrate the current values towards that at a slow enough rate that it appears natural. Regards, Curt. -- Curtis Olson Human Factors Research Lab FlightGear Project Twin Cities cu...@hf... cu...@fl... Minnesota http://www.menet.umn.edu/~curt http://www.flightgear.org |