NAVFLIR hotspot detector - a proposal

I mentioned this already, burried in some thread, but as I just stumbled upon this again in the WIP manual, I'll want to propose something.

 

The WIP manual says: "The real NAVFLIR hotspot detector has more features and options that cannot be simulated on DCS. One of these features is sensibility. The real NAVFLIR hotspot detector will detect ALL temperature differentials which creates many false readings. Unfortunately, it is not possible to simulate such sensibility in DCS, so the hotspot detector is limited to detection of active vehicles (AI or player controlled vehicles). The hotspot detector in DCS will not mark buildings or scenery objects."

 

As we have no temperatures for objects and scenery, I understand that the hotspot detector can only be an approximation. But if the hot spot detector would give false positives and also false negatives, it would not be so "uber" in terms of target detection.

 

Afaik scenery objects can not be queried by a module and so the hotspot detector just has no clue about their presence or absence. But maybe terrain type can be determined? The Viggen seems to be able to distinguish at least water, grass/fields and forrests. Perhaps even urban terrain. I would like to see false positive returns around the borders of different terrain - ofc with only a certain probability. That would emulate to some extend the temperature differences between i.e. a forrest and an open field. Perhaps some more false positives around urban areas? If we want to go overboard with it, take the position of the sun into account and add false positives on water (i.e. reflections of the sun).

 

And ofc not every real object should give a cue in every instance. Maybe the ratio can also depend on the terrain type? (i.e. open fields + truck heated up by the sun, clearly visible > shadowed truck in a dense forrest).

 

What do you think, Razbams? ;-)

I think that DCS World 2.5 will include, if not at release then at some point down the line, a much improved model of infra-red which will hopefully make infra-red cameras realistic for everyone including of course the Harrier.

hrmm ... no objections ... that is a good sign, right? :smilewink:

I mentioned this already, burried in some thread, but as I just stumbled upon this again in the WIP manual, I'll want to propose something.

 

The WIP manual says: "The real NAVFLIR hotspot detector has more features and options that cannot be simulated on DCS. One of these features is sensibility. The real NAVFLIR hotspot detector will detect ALL temperature differentials which creates many false readings. Unfortunately, it is not possible to simulate such sensibility in DCS, so the hotspot detector is limited to detection of active vehicles (AI or player controlled vehicles). The hotspot detector in DCS will not mark buildings or scenery objects."

 

As we have no temperatures for objects and scenery, I understand that the hotspot detector can only be an approximation. But if the hot spot detector would give false positives and also false negatives, it would not be so "uber" in terms of target detection.

 

Afaik scenery objects can not be queried by a module and so the hotspot detector just has no clue about their presence or absence. But maybe terrain type can be determined? The Viggen seems to be able to distinguish at least water, grass/fields and forrests. Perhaps even urban terrain. I would like to see false positive returns around the borders of different terrain - ofc with only a certain probability. That would emulate to some extend the temperature differences between i.e. a forrest and an open field. Perhaps some more false positives around urban areas? If we want to go overboard with it, take the position of the sun into account and add false positives on water (i.e. reflections of the sun).

 

And ofc not every real object should give a cue in every instance. Maybe the ratio can also depend on the terrain type? (i.e. open fields + truck heated up by the sun, clearly visible > shadowed truck in a dense forrest).

 

What do you think, Razbams? ;-)

 

The reason why the increased IR sensitivity for the hot spot detector is not possible to simulate:

1. There are no true hot spots in-game.

2. To check for a hot spot the sensor would have to interrogate everything in its FOV (Field-Of-View), both terrain and objects.

3. This interrogation would take too many computer cycles and will severely hit DCS overall performance, not only on FPS but on everything else and it may even bring down DCS by grabbing all computer resources on mid or low end computers.

4. It is easier to just interrogate the active vehicles list to see if the vehicle is within the sensor's FOV.

 

I talked at length with ED's development team about the hot spot detector. I even proposed to have it also get buildings and some scenery vehicles, but they said that the problem is to do that they would have to go back to interrogate the entire environment again. DCS does not have a scenery list. It does have an "Active Vehicles" list, because those are under DCS and player control, and only active vehicles have the "heat" value that can be used for it.

 

So, in order to create false positives, which are not that difficult to discern anyway, both ED and us must find a way that does not impact DCS performance.

 

The hot-spot detector can only show up to 10 spots on the HUD or MPCD. What the false positives do is fill these 10 slots making possible for you to miss a real target that could not be displayed because the slots were all full.

 

This video shows the hot-spot detector in action. The "V"s are spot with a higher temperature than the surrounding area.

 

Once again another highlight of why the current IR system in DCS needs some work. I mean as far as performance goes Arma 3 has a proper infrared imaging system whereby the actual temperature of objects is factored in, it's a system that works beautifully and it's dynamic - objects starting up get hot, barrels firing get hot, dead bodies etc get cold. And not a single problem with performance, even on low-mid range PCs like I've got. A system like this would solve this issue completely, but it'll likely come with large costs to development for not just IR sensors but also the terrain and other objects.

Edited October 26, 2017 by Northstar98

Does it mean we will always have 10 "V"s on the HUD ?

Does it mean we will always have 10 "V"s on the HUD ?

 

Not necessarily. 10 is the max number that can be displayed. If you have more than 10 vehicles, only 10 will be shown. If you have less than 10, for example 4, only those will be displayed (4 "V"s in the example).

Or just show random V's...... :tomato:

Zeus, thanks for elaborating on this topic. I understand where the challenges are here, but I just want this system to be not too "simple" as it would make it too powerful.

 

We all know how difficult it is to visually detect ground targets - especially in DCS. But with a sensor that now would tell me exactly where they are, with a minimum probability of (100 - x) %, it would spoil the fun greatly, though. A sensor that tells me, "if I point at something, then there IS something" makes it too easy. That is why I am so strongly proposing a random factor to be added.

 

I understand now that it might not be viable to incorporate some sort of contextual elements into that random factor if DCS does not provide such data to a module (i.e. the random factor weighted, based upon the type of terrain). But there are still options that you could consider. Maybe the relative position of the sun? Like if it is behind me, is less likely to produce false positives than if it is in front of me (if makes that makes sense, physically)? Or the time of day (daytime vs. night time?). Or ... dunno, we all could brain storm a bit. :-)

 

And the limit of max. 10 contacts to be displayd are not really a concern here, imo. RL pilots have to deal with that, too. Again, I would not want to have the system always clutter the HUD with 10 contacts, false positives or real ones, all the time. But just enough to make me have to double check. And we should also have false negatives - no contact indicated, although an object IS there.

 

I understand, that adding such a random factor is challaging in regards to "balancing", to make the sensor not totally useless. But on the other hand, a totally "uber" sensor is also not really fun.

Or just show random V's...... :tomato:

 

Better don't show anything than random V's.. :noexpression:

Better don't show anything than random V's.. :noexpression:

Well, tbh most of those Vs seem random in the video posted before. We know they are not, but they look so.

 

An acceptable solution, imo, is adding a certain degree of randomness even for the targets they are actually active (maybe considering the distance from the target or other parameters, that's up to the devs). In other words, I agree on having some false positive if they are generated following a credible criteria.

My solution would be to leave it with a perfect ability to detect vehicles but set the detection range as very short. No more than say 1 or 2 miles.

That would be a bad solution making it even more unrealistic. Unless the real range is 1 or 2 miles.

The reason why the increased IR sensitivity for the hot spot detector is not possible to simulate:

1. There are no true hot spots in-game.

2. To check for a hot spot the sensor would have to interrogate everything in its FOV (Field-Of-View), both terrain and objects.

3. This interrogation would take too many computer cycles and will severely hit DCS overall performance, not only on FPS but on everything else and it may even bring down DCS by grabbing all computer resources on mid or low end computers.

4. It is easier to just interrogate the active vehicles list to see if the vehicle is within the sensor's FOV.

 

I talked at length with ED's development team about the hot spot detector. I even proposed to have it also get buildings and some scenery vehicles, but they said that the problem is to do that they would have to go back to interrogate the entire environment again. DCS does not have a scenery list. It does have an "Active Vehicles" list, because those are under DCS and player control, and only active vehicles have the "heat" value that can be used for it.

 

So, in order to create false positives, which are not that difficult to discern anyway, both ED and us must find a way that does not impact DCS performance.

 

The hot-spot detector can only show up to 10 spots on the HUD or MPCD. What the false positives do is fill these 10 slots making possible for you to miss a real target that could not be displayed because the slots were all full.

 

This video shows the hot-spot detector in action. The "V"s are spot with a higher temperature than the surrounding area.

 

Ok, so the easiest way to "simulate" the false positives, would be to show a random hotspot for every 1-3 active vehicles, by adding x and y coordinates to one out of three objects say between 10-800 m in each direction and have that one fade out after 5 seconds?

 

If we ever hit more than 7 vehicles have a minimum of 2 be false positives and reject 2 real active vehicles.

 

The overall result would be convincingly realistic in appearance, still require to double check and does not require to iterate through any environment objects at all.

 

The only drawback I can see so far, is that we could not estimate from real world objects (buildings, objects, rocks, trees) what are likely false positives.

On the other hand, we can't be sure if the hot spot/contrast isn't just a rock hidden under gras, a pool of water under bushes or whatever...

 

So for me personally it would be a very effective way to simulate the real life behaviour without impacting performance.

Add a few randoms and call it good.

Add a few randoms and call it good.

As Easy as it sounds, it isn't "just add a few randoms".

Think about the 3D area in front of the plane.

Were do you place the randoms? Below the horizon, in range of the FLIR, but what about the position? If you turn are a number of false positives "stored" in memory so they appear again on turning back? How many? How far left and right the HUD do you generate false positives?

 

That's why I supposed to create false positives from the actual identified active vehicles. It would keep a) a reference, b) limit the number of false positives to a manageable amount.

 

Maybe there are even better methods, I can't figure at the moment...

 

But any kind of a couple false positives thrown in the mix without monitoring the whole DCS object list sounds like a good idea.

I'm just happy we're getting this system at all to be honest

I'm just happy we're getting this system at all to be honest

Me too, just want to give Zeus and the team some ideas to improve it as they seem to be focused on the environment objects.

As it is still WIP the best point in time to brainstorm ideas and options...

Add a few randoms and call it good.

The trouble is the FLIR markers aren't just 'random' noise IRL they give important SA that a human can filter.

 

The markers show hot/cold spots and gradients.

 

A row of markers might draw attention to:-

● electricity pylons across a of forest/field

● a railway line/road

● the hot/cold boundary between a lake and shore

● the hot/cold horizon boundary between ground and sky

 

Groups of markers:-

● The sun and/or it's refection in water

● bright reflections off the surface of clouds at altitude

 

A lone marker :-

● a single building,

● hot tin roof

● a sun reflection

● a large vehicle/ship, etc.

 

I think DCS AV-8B FLIR markers should be kept simple and short ranged (2 nm ?) with a risk of non-detection (IIRC FLIR is TOD sensitive, not sure if it's modelled in DCS).

 

A 'neat' exception to the FLIR's short range, might be the Sun (if DCS A2A IR modes can be interrogated) but it might not be worth the extra processing needed?

 

I certainly don't want:-

● random markers over the HUD that distract or obscure for game play reasons.

● To change a planned attack run because of a RNG false marker (IRL a marker is associated with a RL object, so easy to reject/analyse).

● Fly over open sea or in fog/low light and have random FLIR markers appearing on the HUD because of 'reasons'.

 

A-10C TGP, Ka-50 or Gazelle cameras will have range advantage but require more work from the player.

 

Given the different requirements of SP/COOP and PvP, a compromise would be to allow False FLIR markers via Special Options/Mission Editor setting but I can't help think it's overkill due to the limited range of FLIR (all a mission designer needs to do is place a few random vehicles about a target to obscure it?).

Edited October 28, 2017 by Ramsay

The trouble is the FLIR markers aren't just 'random' noise IRL they give important SA that a human can filter.

 

The markers show hot/cold spots and gradients.(...)

That is what Zeus67 meant, it is not possible to do it "realistic".

Unfortunately that means currently we likely get a "unrealistic" 100% accurate "enemy here" marker system.

 

I certainly don't want:-

● random markers over the HUD that distract or obscure for game play reasons.

● To change a planned attack run because of a RNG false marker (IRL a marker is associated with a RL object, so easy to reject/analyse).

● Fly over open sea or in fog/low light and have random FLIR markers appearing on the HUD because of 'reasons'.

Won't happen if you only add false positives to actual targets and keep them limited.

 

Given the different requirements of SP/COOP and PvP, a compromise would be to allow False FLIR markers via Special Options/Mission Editor setting but I can't help think it's overkill due to the limited range of FLIR (all a mission designer needs to do is place a few random vehicles about a target to obscure it?).

I am not sure it is a better idea to require mission designers to crowd the maps with random vehicles to obscure the real targets.

 

I will be very happy if we have any solution that does not mean we have a 100% accurate solution that points us in every target without error...

Unfortunately though the best option means the most amount of work - it is a system that works well in at least Arma 3 but will result in more work. That is a remodel of how infrared works in DCS.

 

In Arma 3 there is an actual dynamic temperature system - object surfaces and in some cases internals actually have a temperature given to them, this temperature is dynamic, so as vehicle engines start-up they get hotter giving better contrast to IR sensors, same goes for barrels when firing or aircraft and vehicles shutting down. Again this takes current weather conditions into consideration. This means we get a more accurate simulation of the Vs

 

Unfortunately this means applying a system to not only IR sensors in DCS but also all of our maps and every single object. Then making it dynamic for weather conditions and states.

 

For me it's a dream FLIR system and in Arma 3 it works beautifully especially having it dynamic.

 

Oh well, just an idea - though a pretty costly one.

I'm all for more accurate 'randoms' placed in the HUD. What I'm more concerned about is nothing being put in but real contacts. After watching that video I can see that the sensor is anything but trustworthy.

 

I also hope they plan to include airborne IR contacts from active AI or Clients.

Unfortunately though the best option means the most amount of work - it is a system that works well in at least Arma 3 but will result in more work. That is a remodel of how infrared works in DCS.

 

In Arma 3 there is an actual dynamic temperature system - object surfaces and in some cases internals actually have a temperature given to them, this temperature is dynamic, so as vehicle engines start-up they get hotter giving better contrast to IR sensors, same goes for barrels when firing or aircraft and vehicles shutting down. Again this takes current weather conditions into consideration. This means we get a more accurate simulation of the Vs

 

Unfortunately this means applying a system to not only IR sensors in DCS but also all of our maps and every single object. Then making it dynamic for weather conditions and states.

 

For me it's a dream FLIR system and in Arma 3 it works beautifully especially having it dynamic.

 

Oh well, just an idea - though a pretty costly one.

As much as many if us would love such a solution, the current limitation is likely the map size.

 

While the tiny ARMA map land mass is a bit smaller than the modeled City of Las Vegas on the NTTR map, this system would still require to query ALL objects on the map like buildings, trees, static cars, signs, traffic lights, bushes, etc. (we exclude the ground and water here) to determine their temperature. So if the view expands over Las Vegas, that would mean a LOT of object temperatures to query...

 

As Zeus said, this would require more CPU load per cycle in the Sim World, which is pretty much a guarantee for one-digit fps...

 

And you would need to store this information in the memory for all objects.

 

What we need here is a balance between fidelity and feasibility to give us realistic results without bogging down performance.

Another factor would be how IR missiles interpret any hotspots. I presume the system would be part of the core software, or are we talking about something exclusive to the AV-8B?

 

It's a toughie, and whatever solution is arrived at, it's bound to have limitations. For me, something that renders an image that gives us a typical set of returns, but doesn't necessarily react to what's on the ground may have to suffice. I dread to think what the hit on processing resources might be if there's an algorithm included to calculate what might or might not throw up false returns.

 

I guess it is just a question of what compromise RAZBAM feel is appropriate.

Another factor would be how IR missiles interpret any hotspots. I presume the system would be part of the core software, or are we talking about something exclusive to the AV-8B?

 

It's a toughie, and whatever solution is arrived at, it's bound to have limitations. For me, something that renders an image that gives us a typical set of returns, but doesn't necessarily react to what's on the ground may have to suffice. I dread to think what the hit on processing resources might be if there's an algorithm included to calculate what might or might not throw up false returns.

 

I guess it is just a question of what compromise RAZBAM feel is appropriate.

^this... A complete overhaul of the IR detection system can only be part of DCS World and thus require a feasible compromise as CPU cycles is a limited resource.