Infrared outdoor scene simulation

radiance-general
1

( PLEASE DO NOT QUOTE MY EMAIL ADDRESS IN THE CLEAR
         AS I DO NOT WANT SPAMMERS TO HARVEST IT FROM THE
         MAILING-LIST ARCHIVE WEB PAGES. Thank you. )

I am currently working on a project that requires high-fidelity,
non-real-time, multi-spectral infrared simulation of large-scale outdoor
scenes. The horizon may be up to 50 km away from the focal point, and
objects of interest may be anywhere from dozens of metres to dozens of
kilometres distant.

I surveyed the available open-source ray tracers, and RADIANCE is the only
suitable candidate I could find. I have read and understood most of the
RADIANCE research papers listed on the web site, gone over the rpict(1) man
page in detail, and have skimmed through all the archived mailing lists. I
have not yet rendered anything nor spent any time reading source code.

I will need to make some modifications to RADIANCE for this project:

* The first modification is a way to specify very irregular subdivision of
the image plane so that the level of detail of and around objects of
interest can be wound up, while "uninteresting" parts of the scene can be
rendered in much lower detail to speed computation. Essentially, this would
provide a kind of selective oversampling.

* The second modification is a different atmospheric propagation model that
can accurately compute infrared as well as visible spectral path emissions
and attenuations between any two points under specified environmental
conditions. The model itself is very extensive, highly validated, and of
military pedigree. The modification to RADIANCE would take the form of an
atmospheric model "plug-in" scheme for rpict(1) and a gensky(1)-like program
that would use the plug-in to compute a sky radiance map and solar/lunar
position+radiance.

I would be more than happy to contribute the results of this work back to
the RADIANCE project, if there is interest. I have two questions:

- Is anyone aware of any similar work that has already been done or may be
  ongoing?

- Does anyone have any comments that might assist me to plan this work, or
  make the outcome more useful for the RADIANCE community? I would like to
  know about any issues such as outstanding performance/accuracy concerns
  with the existing ray propagation model, or limitations to be aware of in
  the rendering of large-scale, outdoor scenes, or any wish-list features
  that may dovetail into this work.

···

--
Andrew Bettison
Acacia Research Pty Limited
< andrewb AT acres DOT com DOT au >

2

my two toughts:

Andrew Bettison wrote:

      ( PLEASE DO NOT QUOTE MY EMAIL ADDRESS IN THE CLEAR
        AS I DO NOT WANT SPAMMERS TO HARVEST IT FROM THE
        MAILING-LIST ARCHIVE WEB PAGES. Thank you. )

you may want to switch on 'conceal email adress from subscriber list' option in the webinterface.

....
I will need to make some modifications to RADIANCE for this project:

* The first modification is a way to specify very irregular subdivision of
the image plane so that the level of detail of and around objects of
interest can be wound up, while "uninteresting" parts of the scene can be
rendered in much lower detail to speed computation. Essentially, this would
provide a kind of selective oversampling.

Either use rtrace as render engine and plug a point/direction generator before and an value-to-image assembler behind that, or start from rpict.

* The second modification is a different atmospheric propagation model that
can accurately compute infrared as well as visible spectral path emissions
and attenuations between any two points under specified environmental
conditions. The model itself is very extensive, highly validated, and of
military pedigree. The modification to RADIANCE would take the form of an
atmospheric model "plug-in" scheme for rpict(1) and a gensky(1)-like program
that would use the plug-in to compute a sky radiance map and solar/lunar
position+radiance.

Gendaylit (Perez sky model) is the only other sky model (gensky substitute) I'm aware of.
Simulating IR viewing through the atmosphere would require fixes/changes/add-ons to the mist material.

....
- Does anyone have any comments that might assist me to plan this work, or
make the outcome more useful for the RADIANCE community? I would like to
know about any issues such as outstanding performance/accuracy concerns
with the existing ray propagation model, or limitations to be aware of in
the rendering of large-scale, outdoor scenes, or any wish-list features
that may dovetail into this work.

Since Radiance uses a bounding box for the scene, a very large scene probably requires adjusting some default rpict parameters (e.g. -ar).
Otherwise, most internal variables are 'double'.

-Peter

···

--
pab-opto, Freiburg, Germany, www.pab-opto.de