Normals tables | Olaf van der Spek | 22:41 09-11-2002 | | |
Re: Normals tables | flyby | 20:05 11-11-2002 | | |
> At http://xcc.tiberian.com/documents/ you can find a document written by Ghost that contains normals tables.
Normally you need 8 times a 256 table. One for each facing direction. :)
The normal table presented here only has identified 6 directions. If you take my RA2 normal table I made for VSE r1.6B3, you'll see i got up to 26 fairly easy....
If you dig deeper and deeper into the normal problem at certain point you'll understand that you need a spatial model to relate the numbers to a position. I got much farther then what is presented here, only to realize that beyond a certain point this is a dead end.
you can't use a 16x15 or 16x16 spherical model for the following reasons :
-that type of spatial model has Poles, where there is no surface at the poles. This does not correspond with empiric observation of WW models.
-another consequence of such a model is that the surfaces get smaller and smaller the closer they are to the poles.
Again observations on WW voxels indicate otherwise...
After that conclusion i shifted my attention to geodetic spatial models, in the hope to find a model that corresponds with the voxelnormals. I couldn't find any existing object that matches 244 normals. For a while i did put all hope on a model that had 256 faces. But my excel normal tables obtained from my test slabs early this year showed that normal beyond 244 do NOT react as all other normals. Their color shifting is even less then the 2 top/bottom normals.
What caused a breakthrough in my research is the understanding that we are all looking at it the wrong way :
normals are values that correspond with the angle of polygons of a 3D mesh object. Because of our common knowledge of mathematics, we're all looking for normal surfaces as they give an indication of the angle of the surface. Here is were we all go wrong: myself and all other people are looking for nicely defined surfaces.
BUT
What we really need is a direction...
Now, follow that logic...what defines a direction?
2 points...
Here we have it : the normal direction is defined by 2 points: one is at the centre of the object, the other is...
a VERTEX on a spherical model. And NOT a surface...
SO
i went looking for a spatial model with 244 vertices and ... found one right in 3dsmax.
The spherical model I’m talking about here is a geodesic Tetra based sphere, level 11.
I've already started developing some algorithms that are capable of comparing the normal directions of a polygon with one of the predefined 244 directions.
My real problem is that i haven't had any free time since 3-4 months. I need at least a week (60-90hrs) to dig into the topic again and continue my programming. I've come to the conclusion that if my professional work keeps increasing like this i wont have any time to finish it...
If I manage to get one day free, I’ll do my best to publish all this into a paper, so Olaf can store it among his documents...