Re: [Algorithms] portal engines in outdoor environments

[Mark W. <mwi...@cy...>]

----- Original Message -----
From: "John Ratcliff" <jra...@ve...>
To: <gda...@li...>
Sent: Friday, August 18, 2000 4:19 PM
Subject: RE: [Algorithms] portal engines in outdoor
environments


>
> >at large polygon walls. The eye-to-vertex algorithm tells
me the wall is
> not
> >visible because I can't see its vertices. I don't know
what to do about it
> >yet. Any ideas?
>
> Send more rays.  For large polygons tesselate them down to
some minimum
> threshold size and send rays to them.
>
> John
>

Assuming all this work is done as a pre-process, wouldn't it
be easier to simply render the scene and read back the frame
buffer?  To be more specific:

1) Assign each node some unique RGB color.
2) Render all polygons in a node using its unique RGB color.
3) Read back the frame buffer and record all unique RGB
values in the image.  This will give you all nodes visible
from this node.  The Z-Buffer will take care of eliminating
any invisible nodes.
4) If you do this with 6 images/orientations (like those
used to make skyboxes) you will cover the entire view.

The main problem (which also exists in John's method) is
that you don't know where to place the camera inside the
node before you render the images or do your ray-casting.  I
guess you would need to repeat all this from random
locations within a node to make sure you catch all possible
cases.  IIRC, the method I described above was used for
Rogue Squadron on the N64.  They used some hi-end SGI
workstation to render thousands of images and analyze them
to determine visibility.

-Mark