Generating a heightmap 2004-02-24 - By Danny Kodicek
Back
It sounds like you're overworking it, really. If they're just boxes, then
determining which tiles each box overlaps is fairly trivial, and then it's
simply a matter of calculating the maximum height of each box that overlaps
a particular tile. I assume they're flat to the floor - if they could be
tilted then it's a bit more complicated (but not obscenely so)
I'd probably do it as you build the level: each time you place a box,
calculate which tiles it overlaps and adjust their max_height property
accordingly.
Danny
-- -- Original Message -- --
From: "Ian Thomas" <ian@(protected)>
To: <dir3d-l@(protected)>
Sent: Tuesday, February 24, 2004 1:17 PM
Subject: [Dir3d-l] Generating a heightmap
> Hi guys,
>
> I've been wrestling with this one for a while, and wondered if anyone
> had any inspiration.
>
> The setup: I've got a finite plane (the 'floor') of 240x240 at y=0,
> scattered around which may be an arbitrary number of solids (the
> contents of the room). The solids are just #box primitives, scaled,
> rotated and translated but still essentially boxes.
>
> I've divided the room into tiles of 12x12 for pathfinding purposes.
>
> Basically I need to establish what the highest point is within in each
> tile. (Note this may not be a vertex, as objects could cheerfully span
> tiles without having a vertex within a tile.)
>
> The routine I'm building is for precalcing the heightmap, so doesn't
> need to be blistering - however, I'd like it to be pretty fast as my
> room editing utility needs to rebuild on a regular basis. :)
>
> Approaches I've tried:
>
> Brute force:
> -- ---- ----
> - Apply a meshdeform to each object.
> - For each tile:
> - Discard objects whose bounding sphere is nowhere near.
> - For each valid object, use Sutherland-Hodgeman clipping on each
> face. If any of the face is inside, use the clipped vertices and take
> the max height.
>
> This seems to work very well, but is exceedingly slow; add to that I get
> occasional anomalies and wonder if my S-H clipping implementation has
> holes in (if anyone can point me at a good bit of sourcecode in whatever
> language then I'd appreciate it).
>
> Collision:
> -- ---- --
> - Turn on collision detection for the relevant objects
> - For each tile:
> - Move a tile-shaped solid upwards from y=0 upwards until it stops
> colliding. That's the max height.
>
> This works okay, is slow, requires updatestages and sleeps to make the
> collision callback work, and the collision doesn't seem to be 100%
> accurate.
>
> Raycasting:
> -- ---- ---
> - For each tile
> - Cast a network of angled rays downwards towards the floor (within
> the volume of the tile) - the heighest point hit by a ray is the max
> height for the tile.
>
> This lacks precision - sometimes very thin blocks escape the net. :( To
> catch those, you need more rays. The more rays, the slower it is.
>
> Things I thought of but haven't tried:
>
> Additive:
> -- ------
> - Reverse the process, and actually build the heightmap as I build the
> map - for each object, work out which tiles it lies in and set heights
> appropriately.
>
> To be honest, I gave up on the maths for this in favour of trying the
> other things. It sounds plausible but I still need to do lots of
> clipping. :(
>
> Any blindingly obvious solutions that I've missed? If not, I think next
> step is to look at the additive solution using Sutherland-Hodgeman
> *sigh* but it ain't pretty. :)
>
> TIA,
> Ian
>
> __ ____ ____ ____ ____ ____ ____ ____ ____ ____
> Dir3d-l mailing list
> Dir3d-l@(protected)
> http://nuttybar.drama.uga.edu/mailman/listinfo/dir3d-l
__ ____ ____ ____ ____ ____ ____ ____ ____ ____
Dir3d-l mailing list
Dir3d-l@(protected)
http://nuttybar.drama.uga.edu/mailman/listinfo/dir3d-l
|
|
