Jump to content

Thermal Vision Improvements From A Physicist


127 replies to this topic

#41 Shamous13

    Member

  • PipPipPipPipPipPipPip
  • 684 posts
  • LocationKitchener, Ont.

Posted 04 April 2013 - 05:10 AM

This is the best post I've read in a long time. I hope it grabs their attention.

Edited by Shamous13, 04 April 2013 - 05:11 AM.


#42 CnlPepper

    Member

  • PipPip
  • Knight Errant
  • 41 posts

Posted 04 April 2013 - 05:24 AM

Jay, the images are mockups with hand painted thermal signatures, they'd look quite a bit better if generated from real meshes. Also please consider the closest mech in the images is a tiny light mech, larger mechs would be clearer. The pixel resolution for the later images is fairly consistent with a medium end commercial thermal imaging sensor.

Despite being pixelated I'm comfortable that the mechs are fully targetable, I think I would stand a good chance of hitting the torso of the mech at 700m. Just because you can't make out a head doesn't mean you can't do it damage!

Something that will help is that when the scene is animated, ie not a static image, you will be able to infer much more information.

The benefit of the reduced resolution approach is that we retain the potential to spot long range targets, but it is now a much greater challenge to actually hit them, especially > 1000m. Sniping in the game with the old heat vision was very easy. I have a few "sniper" mechs and I would simply leave thermal mode engaged on any map. The devs have stated that they want the different vision modes to be dominant on different maps/times of day. I believe this proposal would better meet that critera than the current updated themal mode (which is extremely unrealistic and suffers some very unintuitive quirks).

Edited by CnlPepper, 04 April 2013 - 07:28 AM.


#43 Karl Split

    Member

  • PipPipPipPipPipPipPip
  • 727 posts

Posted 04 April 2013 - 06:20 AM

PGI, please give the OP a job, if he's already got one pay him more!

Great suggestions, I love the idea of module upgrades for em too, if someone wants to be a sniper it'll give him the modules to do it, but force him not to take other useful modules.

#44 Rippthrough

    Member

  • PipPipPipPipPipPipPipPip
  • Mercenary
  • 1,201 posts

Posted 04 April 2013 - 06:28 AM

The ones with the doubled pixel resolution look absolutely perfect to me, great work, much better than this hash up system we have at the minute (which only really cuts off the 700-1000m mark, beyond that you can see fine).

Nice work, I'd definately like this to be considered as a replacement for the current thermal situation.

#45 frag85

    Member

  • PipPipPipPipPip
  • 141 posts

Posted 04 April 2013 - 07:09 AM

View PostCnlPepper, on 02 April 2013 - 02:41 PM, said:


1) Render the thermal scene to a low resolution texture eg 192x108 pixel and then stretch it across the view window. In practice this will need to be oversampled to prevent aliasing, so say a 768x432 buffer and resample to 192x108.



Nice write up. The only suggestion I have is to have the resolution a percentage of the user's resolution to accommodate everyone in an equal manner. Not everyone plays at the same resolution. People with higher resolutions would be at a disadvantage and anyone with a lower res, an advantage.



Edit: Could range also be limited by the way a protective lense over the sensor works? Or even just the sensitivity of the sensor? More so on the latter. I'm looking at this from an "if you have 255 levels of grey to work with" angle. On an older thermal imager I have used with a fire department (Scott Eagle Imager 6, IIRC) after a few hundred feet everything washes out. This has always made me wonder because you can see very detailed images up close. Even to the point that (pardon the immaturity here) when someone farts, you can see a puff of white.

On hot maps, it would be much more difficult to see cool mechs, and hot mechs would be less distinguished.

Edited by frag85, 04 April 2013 - 07:26 AM.


#46 jakucha

    Member

  • PipPipPipPipPipPipPipPipPip
  • Knight Errant
  • 2,413 posts

Posted 04 April 2013 - 07:10 AM

Since they're going to tweak thermal as it's not in its final state, I hope it looks a bit more like this when they do.

#47 Dakkath

    Member

  • PipPipPipPipPipPipPipPip
  • 1,980 posts
  • LocationG-14 Classified

Posted 04 April 2013 - 07:19 AM

IMO, Much too pixelated to be used in production, people would freak out. But, I like the idea of what you're after.

#48 CnlPepper

    Member

  • PipPip
  • Knight Errant
  • 41 posts

Posted 04 April 2013 - 07:38 AM

Dakkath, I'm interested to know why you think people would freak out? It isn't any different to any other post processing filter. I don't see it being difficult to learn how to use - people got use to the old thermal view just fine.

Frag85, I suspect the reason the distance was blured on your camera was due to it having a short focal length ie the limitation was in the lens system rather than a limitation of the sensor itself. It is something that could be added to the simulation but it would require extra post processing to produce which would potentially impact frame rates for lower end users.

#49 Mordhar

    Member

  • PipPipPip
  • Big Brother
  • Big Brother
  • 69 posts
  • LocationChelyabinsk, Russia

Posted 04 April 2013 - 08:21 AM

Very good idea. But allow me to add some of my own thoughts.

1) I suggest to use thermal (and, maybe, night) vision modes in a way close to “advanced zoom” module. A small (2/3 or 1/2 of the screen) and highly pixilated picture-in-picture image.
Pixilation of “thermal imaging” may (or may not) hamper player’s ability to navigate and maneuver his/her mech. And zone of normal, non-altered image at the edges of screen will make this problem less prominent. At the same time, narrower “thermal image” will force to use it as supplementary mode, not “always-on easy-mode”.

2) Also add some “trails” or “motion blur” to objects, to simulate lower reaction time of IR sensors. It will make extreme-range sniping in thermal mode more difficult. Thus, excessive pixilation will be unnecessary.

3) As for “advanced thermal imaging” module – it can be ability to use color-coded (old) thermal mode instead of grayscale.

In any case, current implementation of thermal vision needs to be reworked. White snow in Alpine and dark ground in Caustic not only looks unrealistic, but also goes against the concept of “different modes for different situations”.

View PostDakkath, on 04 April 2013 - 07:19 AM, said:

IMO, Much too pixelated to be used in production, people would freak out. But, I like the idea of what you're after.


“Advaced” zoom module not only pixilated, but also have magnified “noise” overlay on it. And it is already in game.

Edited by Mordhar, 04 April 2013 - 08:24 AM.


#50 jay35

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,597 posts

Posted 04 April 2013 - 08:25 AM

It's good CnlPepper, just not quite that pixelated. I appreciate what you're aiming for, it's just perhaps a bit too stringent for the audience.

#51 The Black Knight

    Member

  • PipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 160 posts

Posted 04 April 2013 - 09:43 AM

Its my opinion that all progress on creating new content should be paused, and have the entire PGI team focus on fixing the core game mechanics. I have found myself playing less and less since the jager patch, and even rage quit yesterday, because all the glitches and bugs detract from the experience. Before the jager patch the game was an near perfect as it had ever been and I truly miss that. We all want MWO to have as many new players as possible, but if those of us who have been playing since closed beta are no longer wanting to play how can we expect NEW players to want to play? I love battletech and MechWarrior and I hate to see the game heading backwards. I don't want a new assault mech if I can barely stand to play with the ones I have. Please, PGI, focus on fixing the game so many of us love.

#52 MangoBogadog

    Member

  • PipPipPipPipPipPip
  • Liquid Metal
  • Liquid Metal
  • 377 posts
  • Google+: Link
  • Facebook: Link
  • LocationUK

Posted 04 April 2013 - 01:36 PM

Current thermal is not too great, I like CnlPepper's solution more.

Edited by MangoBogadog, 04 April 2013 - 01:43 PM.


#53 Dakkath

    Member

  • PipPipPipPipPipPipPipPip
  • 1,980 posts
  • LocationG-14 Classified

Posted 04 April 2013 - 02:21 PM

View PostCnlPepper, on 04 April 2013 - 07:38 AM, said:

Dakkath, I'm interested to know why you think people would freak out? It isn't any different to any other post processing filter. I don't see it being difficult to learn how to use - people got use to the old thermal view just fine.

Frag85, I suspect the reason the distance was blured on your camera was due to it having a short focal length ie the limitation was in the lens system rather than a limitation of the sensor itself. It is something that could be added to the simulation but it would require extra post processing to produce which would potentially impact frame rates for lower end users.



Because right now its rather clean, and usable. If you were to change it to that pixelated-form it wouldn't go over to well with the people. At least this is my opinion. I know I would rather have a clean, usable interface than a truer-to-form pixelated mess.

@Mor

Quote

“Advaced” zoom module not only pixilated, but also have magnified “noise” overlay on it. And it is already in game.


I know, but they have plans to fix this, because it does indeed suck. :rolleyes:

#54 CnlPepper

    Member

  • PipPip
  • Knight Errant
  • 41 posts

Posted 04 April 2013 - 03:37 PM

I've been having another look at the shaders used by the game and I think it would be relatively easy to implement a reasonably good thermal vision mode using the following recipe. This recipe uses the diffuse textures as a proxy for the emissivity and manipulates the illumination levels using a simple formula that depends on a passed heat variable - basically following the approach I used to produce the mock up images:

For each shader (terrain, mech etc...) the fragment shader simply needs:

1) a temperature constant passed for each object with range [0,1] which corresponds to the temperature range [-55, 200]: TEMPERATURE

2) the diffuse texture: DIFFUSE

In pseudo code:

// constant to tweak effect
alpha = 0.25
 
// average diffuse value
v = (DIFFUSE.r + DIFFUSE.g + DIFFUSE.:rolleyes: * 0.333333333
 
// compress range to [0, alpha] to simulate range of "emissivities/temperature variations", with alpha being maximum value that corresponds to temperature passed to the shader
v = v * alpha
 
// shift value by passed temperature and set fragment colour
OUT.rgb = saturate(TEMPERATURE + v - alpha)

Obviously this can be optimised, I laid it out for clarity.

For the terrain the temperature passed to the shader would just be the temperature of the map.

For mechs the temperature passed is:

map ambient + mech_temp_offset + mech_temp_scaling * mech_ingame_temp

where the mech_temp_offset sets the minimum temp a cool mech is above ambient and mech_ingame_scaling is obviously a scaling relating the in-game mech temperature percentage to an increase in visible mech temperature. Both constants being tuneables for balance.

For the sky the diffuse should be near zero... maybe it is just easier to set it to zero as the sky is cold.

A post processing step would then follow to either pixelate (the preferred choice) or Guassian blur the heat image by an amount determined by game balancing.

EDIT: it may also be worth adding a diffuse texture multiplier that is defined in the objects material file and passed to the shader, this could be used to brighten diffuse textures for objects with darker diffuse textures. The mech textures may need this tweak.

If you use the above approach you should be able to produce a good effect with only minimal asset tweaks.

Edited by CnlPepper, 04 April 2013 - 04:39 PM.


#55 CnlPepper

    Member

  • PipPip
  • Knight Errant
  • 41 posts

Posted 04 April 2013 - 03:54 PM

Incidentally as a few people don't like the idea of pixelation, here are some mockups using a Guassian blur instead. As you can see, this is effectively what the old thermal vision was and it is very easy to estimate where the target mechs are even at huge ranges as the peak brightness corresponds almost exactly to the heat centre of the mech. This is the reason I suggest a pixelated view as it makes it inherently difficult to judge where the heat centre of a distant mech is. I hope my reasoning is now clear - the pixelation allows the devs to balance accuracy with range.

Posted Image
Posted Image

Edited by CnlPepper, 04 April 2013 - 03:57 PM.


#56 WVAnonymous

    Member

  • PipPipPipPipPipPipPipPip
  • Wrath
  • Wrath
  • 1,691 posts
  • LocationEvery world has a South Bay. That's where I am.

Posted 04 April 2013 - 04:20 PM

View PostCnlPepper, on 04 April 2013 - 03:54 PM, said:

Incidentally as a few people don't like the idea of pixelation, here are some mockups using a Guassian blur instead. As you can see, this is effectively what the old thermal vision was and it is very easy to estimate where the target mechs are even at huge ranges as the peak brightness corresponds almost exactly to the heat centre of the mech. This is the reason I suggest a pixelated view as it makes it inherently difficult to judge where the heat centre of a distant mech is. I hope my reasoning is now clear - the pixelation allows the devs to balance accuracy with range.

Posted Image
Posted Image


I think either way it is "shoot the white spot" but I do think the Gaussian is more esthetically pleasing.

#57 Undecided

    Member

  • PipPip
  • 25 posts

Posted 04 April 2013 - 06:50 PM

Maybe I'm just weird, but I actually prefer the pixilation to blurring. Anyway, I love the concept, it gives thermal a role to play that varies depending on the map without turning it into the go to vision mode like the old thermal. The current thermal view just looks like normal vision in gray-scale with fog.

#58 TopSpeed

    Member

  • Pip
  • 13 posts

Posted 04 April 2013 - 08:05 PM

While I certainly like the idea, the mock-ups look like it would make it too hard to see the terrain sufficiently to navigate. Having used thermal and night-vision, I do not believe that would normally be the case. Besides, with this being a game, I would rather have fun than worry overly about realism. Some is good; too much is more like work.
Regardless, an excellent, well thought-out post and series of responses. Kudos.

#59 p4g3m4s7r

    Member

  • PipPipPipPipPip
  • Little Helper
  • 190 posts

Posted 04 April 2013 - 09:06 PM

View PostDakkath, on 04 April 2013 - 07:19 AM, said:

IMO, Much too pixelated to be used in production, people would freak out. But, I like the idea of what you're after.

View PostDakkath, on 04 April 2013 - 02:21 PM, said:



Because right now its rather clean, and usable. If you were to change it to that pixelated-form it wouldn't go over to well with the people. At least this is my opinion. I know I would rather have a clean, usable interface than a truer-to-form pixelated mess.

@Mor

I know, but they have plans to fix this, because it does indeed suck. :P


As the OP points out, these are commercially fielded systems. Meaning they are highly usable. I think pixelation is the best way to make sure that you can't tell where the center of the mech is from too far away.

You could have a higher resolution than the OP suggests, and use some sort of heat halo to make it more difficult to see the heat center. Gaussian blur is actually much less usable to me. The mind naturally looks for hard edges, and when it can't find them assumes your eyes are out of focus and tries to fix it, which produces eye strain. Eye strain needs absolutely should not be a balancing factor.

You could also arbitrarily remove the ability to zoom in. IR optics (depending on what wavelength is being used) can't use the exact same focusing optics as EO (visible light), so you could argue the lack of zoom ability is because you only have EO zoom optics. Adding in a software zoom is not optimal, because even the broken extended zoom we currently have is really helpful. Being able to shoot at the center of a big pixel is still really useful.

#60 Karyu

    Member

  • PipPipPipPipPip
  • Knight Errant
  • Knight Errant
  • 148 posts
  • LocationHawaii

Posted 05 April 2013 - 02:52 AM

While I appreciate the effort and thought that was put into this post, I'm going to have to strongly disagree with the suggestion. I am not even going to presume to know what sort of technology you have worked with as a physicist but the current incarnation of thermal is actually very similar to modern day composite infra-red. FLIR (Forward Looking Infra-Red) for anyone with military/intelligence experience. Perhaps they should simply rename the vision mode to eliminate confusion?

Pixelation, as with any digital viewing device, is relevant to the quality of the equipment. Arbitrarily adding pixelation to supposed military grade equipment doesn't sit well with me when, as I said, the current version is reasonably realistic.

However, I wouldn't oppose pixelation in the zoom modes. The reason for this being; the game actively renders the environment at the correct resolution when you zoom in, resulting in crystal clear focus in any vision mode at any range. Any equipment I have ever worked with isn't capable of enhancing imagery on the fly and results in a mild, but noticeable loss of resolution as you zoom in on objects.





4 user(s) are reading this topic

0 members, 4 guests, 0 anonymous users