Dear all,
I posted a while ago about NVIZ making "spiky" vectors when draping over a surface. Hamish Bowman suggested that it might be simply because I'm draping a circle over an irregular surface, so that seen from the side it is distorted, but I don't think that's it. I remain baffled by this behavior, and wonder if someone might see what the problem is by looking at some images. I've posted what I'm trying to do at:
http://www.wisc.edu/arth/sardis/nviz.spiky.jpg
There should be no linear features on this -- only circles. The linear "spikes" radiating out from the circles are just not there in the data, but are a bug or something in nviz.
I think the issue has to do with resolution of the DEM vs. length of vector. I'm using a 90 meter DEM, draping over that a 30 meter resolution TM image, and then draping vectors which consist of "circles" -- i.e. 360 short lines in a circle. The mean diameter of the circles is about 43 meters, so the lines are much shorter than the cell size of the DEM. I'm setting the region resolution to 90 before running NVIZ. When I set the polygon resolution within NVIZ to 1 or 2 for a high-res image, the vectors get all spiked out. When I set the polygon resolution to something much higher, however, like 100 or 500, the vectors are drawn correctly as circles.
These are illustrated here:
polygon resolution 2: http://www.wisc.edu/arth/sardis/test.2res.jpg
polygon resolution 5: http://www.wisc.edu/arth/sardis/test.5res.jpg
polygon resolution 100: http://www.wisc.edu/arth/sardis/test.100res.jpg
My temporary solution was to draw the image twice, once with the dem and satellite overlay, and then a second time with the polygon resolution at 500 and drawing only the vectors, and then combining the two images in photoshop. This isn't a perfect, solution, though, because the vectors drawn with the high polygon resolution are actually drawn in different places - apparently changing the polygon resolution changes the effective surface that the vectors are draped on, so they appear sometimes above and sometimes below where they were in the high-res rendition, so the circles end up wrongly placed. I overlaid two images made at 2 and 500 resolution to check out what was happening here -- the blue vectors are at 500, the red at 2
http://www.wisc.edu/arth/sardis/ vectors.2and500.jpg
This doesn't happen with all data -- I have another, 2 meter DEM of a smaller area, and draping even very detailed vectors over that causes no problem. So I'm bamboozled. Does anyone have any idea of how to achieve this?
I did try creating a raster from the vector map, overlaying that over the TM image, and then simply draping that over the DEM. The results just aren't as clean, though, and if possible I'd like to do it with a vector overlay. However, if there's some unfixable bug in NVIZ or some factor I'm forgetting, I'll try it the other way.
I'm running GRASS 5.0 on Mac OSX.
many thanks,
Nick Cahill
Dept. of Art History
UW-Madison