From: Ralf G. <rge...@kn...> - 2008-01-08 09:23:22
|
Hi! Frederic Bouvier wrote: > Selon Curtis Olson : >> My gut feeling is that once you get up (or down) into the latittudes where >> the tiles get significantly skinny, the resolution of the available data >> drops of significantly. We really don't have a per-tile triangle budget >> anyway. The only place where I see this making a difference is the >> concentration of terrain elevation points would increase, but this is up in >> an area where we only have very low res terrain data anyway. SRTM drops out >> beyond +/- 60 degrees latitude. > > I was thinking about the parameter we pass to "Terra" to simplify the initial > grid. IIRC, this parameter is always the same, leaving all *.arr.gz files with > the same number of vertices. Well, from my experience the elevation grid is not the main driver in terms of triangles. In terms of triangles per area, probably the linear features (roads, rivers, railways, etc.) are much more intensive than the basic elevation grid provided by Terra. Further, the width of the tiles is not necessarily proportionally related to the number of triangles, as vegetation and buildup clearly drop in the latitudes farther north and south. If we really can now afford having tiles of different size in terms of linear measure, I would very much favour a regular grid in the latitude-longitude space. That would leave us with only slight changes in SGBucket with only very small adaptations necessary in the rest of the code (FlightGear, TerraGear), assuming that it uses only the interface of SGBucket as it should and not making any further assumptions about its workings. And it would solve the TerraGear neighbour-problems Curt mentioned earlier in the thread. The problems we have regarding the degeneration of quad-tiles at the pole to actually triangular tiles and the fact that the earth surface is sperical and infinite and not flat and finite would remain, but could be worked around much more easily with such a simple and before all consistent grid definition. Cheers, Ralf |