Camouflage

Okay, so right now, we need to discuss camo. It's almost weirdly important in the natural world, and should be mimicked if possible. After all, you're far more likely to survive if you avoid the fight rather than expending the resources and energy required to win one.

The key here is color comparison, and the computer's ability to notice something out of the ordinary in an environment. That is very, very, very difficult to program, as it effectively requires the computer to start to generalize, an action that is inherently animal.

Discuss.

Tenebrarum wrote:

Okay, so right now, we need to discuss camo. It's almost weirdly important in the natural world, and should be mimicked if possible. After all, you're far more likely to survive if you avoid the fight rather than expending the resources and energy required to win one.

The key here is color comparison, and the computer's ability to notice something out of the ordinary in an environment. That is very, very, very difficult to program, as it effectively requires the computer to start to generalize, an action that is inherently animal.

Discuss.

Colours should be easy enough (don't hurt me programmers).

I'm thinking something like:

Find random point on edge of creature (could be tough, i suppose)
Get colour, store as ColourOrg1
Get colour of object behind creature from hunter's line of sight, store as ColourEnviro1
// Have a few of these that choose different edges or different coloured edges, use convention ColourOrg[x] and ColourEnviro[x]
Compare ColourOrg[x] and ColourEnviro[x]
// If done in hex values, this should be as simple as a subtraction, right?
If difference is large = easy to detect, if difference is low = hard
// probably use some sort of scaling in the above line, to allow for a wide variety of detectability

And here's to hoping i didn't just make roadkill cry

Tenebrarum wrote:

Okay, so right now, we need to discuss camo. It's almost weirdly important in the natural world, and should be mimicked if possible. After all, you're far more likely to survive if you avoid the fight rather than expending the resources and energy required to win one.

The key here is color comparison, and the computer's ability to notice something out of the ordinary in an environment. That is very, very, very difficult to program, as it effectively requires the computer to start to generalize, an action that is inherently animal.

Discuss.

I could have sworn this thread already existed. No matter.
For camouflage (wonderfully difficult to spell), we will need to know the basic colors of the environment. This should be easy, since we'll already know them because they will already exist as textures.

So, for animal camouflage, we have a few different options. Animals could have:
the same colors as their surroundings
the same shape as their surroundings
the same texture (exact color scheme) of their surroundings

This is all going to depend on the textures used for rocks, ground coverings and plants in their biome.

Since plants will have to have a procedural texture, I say that animals should be able to get that exact same procedural texture on their bodies. This is the simplest way to simulate camouflage visually for the player. To actually simulate it in the game, the animal simply gets a high chance of not being sensed by the vision of a predator in their presence.

Another way to simulate camouflage (best for large animals) is to simply mimic the colors of their environment. This means finding out what colors are in their environment. The best way to do this is to take a picture from a point inside their environment, exclude the skybox, and analyze what different colors are in the picture. these colors then help the animal blend in when they are on its body.


Click to go to galleries of animal camouflage

I'm not a huge expert on programming yet, so I don't know how hard this idea would be to implement. But here's an idea that doesn't take the computer to actually try to differentiate between colors in game like Alaska's idea.

Maybe environment colors could be placed in overall groups, (light green, green, dark green, etc). Plant color would be the main determining factor, although in areas like deserts, soil could play a part too. Then, when creatures evolve, the computer could check if they are in the same color group as the environment and determine a camouflage variable based off of it. For example, say the environment is a forest, and the overall color is a dark green. A yellow creature exists, and evolves into a shade of light green variation. While the creature doesn't have perfect camouflage, it's camouflage variable would move higher on the positive side of the scale. The variable would then determine how visible it is to predators. Creatures closer to the environment color would have less of a chance of being seen than ones farther from it.

As for patterns such as stripes, the particular biome the creature lives in could be the determining factor, since creatures that live in areas with lots of light and dark areas such as forests are more likely to have patterns, while creatures in open areas like deserts tend to be more flat in color. This could be added to the camouflage variable, with creatures in variable light environments getting a bonus from having a pattern.

This would probably not be as realistic as the ideas already posted here (I'm leaning towards scio's idea), but may be easier to implement.

Quote :

Maybe environment colors could be placed in overall groups, (light green, green, dark green, etc). Plant color would be the main determining factor, although in areas like deserts, soil could play a part too. Then, when creatures evolve, the computer could check if they are in the same color group as the environment and determine a camouflage variable based off of it. For example, say the environment is a forest, and the overall color is a dark green. A yellow creature exists, and evolves into a shade of light green variation. While the creature doesn't have perfect camouflage, it's camouflage variable would move higher on the positive side of the scale. The variable would then determine how visible it is to predators. Creatures closer to the environment color would have less of a chance of being seen than ones farther from it.

As for patterns such as stripes, the particular biome the creature lives in could be the determining factor, since creatures that live in areas with lots of light and dark areas such as forests are more likely to have patterns, while creatures in open areas like deserts tend to be more flat in color. This could be added to the camouflage variable, with creatures in variable light environments getting a bonus from having a pattern.

That's quite simple, and would work quite well. We can definitely classify different shading and pattern types by environment, and that information can be packaged with the biome definition.

Three words: alternate sensory systems.

Tenebrarum wrote:

Three words: alternate sensory systems.

Meaning electric/sonic/olfactory/etc?

Tenebrarum wrote:

Three words: alternate sensory systems.

Regardless of what kind of senses a critter is using, there will still be some kind of graphical output of the environment, yes? This means everything must output some kind of visible texture (unless you guys have come up with smell-o-vision while I've been away). Camo based on visual/infrared/UV can all still be compared to the environment. I'm envisioning a comparison between the color data (as found in graphics editing programs) of a creature, and that of the environment.

I'm on a phone, so I can't throw pictures up, but I know color value analysis tools exist in Photoshop. Perhaps some of the other visually-inclined people here know what I'm talking about. What I'm not aware of is the under-the-hood demand needed to replicate those tools.

Redstar wrote:

Tenebrarum wrote:

Three words: alternate sensory systems.

Regardless of what kind of senses a critter is using, there will still be some kind of graphical output of the environment, yes? This means everything must output some kind of visible texture (unless you guys have come up with smell-o-vision while I've been away). Camo based on visual/infrared/UV can all still be compared to the environment. I'm envisioning a comparison between the color data (as found in graphics editing programs) of a creature, and that of the environment.

I'm on a phone, so I can't throw pictures up, but I know color value analysis tools exist in Photoshop. Perhaps some of the other visually-inclined people here know what I'm talking about. What I'm not aware of is the under-the-hood demand needed to replicate those tools.

Sonar.

Tenebrarum wrote:

Redstar wrote:

Tenebrarum wrote:

Three words: alternate sensory systems.

Regardless of what kind of senses a critter is using, there will still be some kind of graphical output of the environment, yes? This means everything must output some kind of visible texture (unless you guys have come up with smell-o-vision while I've been away). Camo based on visual/infrared/UV can all still be compared to the environment. I'm envisioning a comparison between the color data (as found in graphics editing programs) of a creature, and that of the environment.

I'm on a phone, so I can't throw pictures up, but I know color value analysis tools exist in Photoshop. Perhaps some of the other visually-inclined people here know what I'm talking about. What I'm not aware of is the under-the-hood demand needed to replicate those tools.

Sonar.

All we have to do is compare normal maps/bump maps with ones from the environment. That would give you physical surface texture camo.

Tenebrarum wrote:

Sonar.

Sonar is more a sense that helps the predator find its prey. Maybe going off my idea sonar could be a bonus for predators that lowers the camouflage level of its prey when they're in the vicinity of it. But otherwise I think scio pretty much said it.

However, I just thought of a problem. Using a physical surface map doesn't apply to flying creatures using sonar that hunt down other flying creatures (example bats after insects). Could it just be that the prey in this situation wouldn't have much camouflage, besides if they're in or outside the predator's sonar range? Or perhaps insects have a way of hiding from bats that I don't know about?

tklarenb wrote:

Tenebrarum wrote:

Sonar.

Sonar is more a sense that helps the predator find its prey. Maybe going off my idea sonar could be a bonus for predators that lowers the camouflage level of its prey when they're in the vicinity of it. But otherwise I think scio pretty much said it.

However, I just thought of a problem. Using a physical surface map doesn't apply to flying creatures using sonar that hunt down other flying creatures (example bats after insects). Could it just be that the prey in this situation wouldn't have much camouflage, besides if they're in or outside the predator's sonar range? Or perhaps insects have a way of hiding from bats that I don't know about?

Flying creature attempting to avoid sonar would have to use the biological equivalent of RAM (radar absorbing material). Their body covering would scatter and absorb sound waves.
also, If you don't mind, could you downsize your avatar a bit. I like it but it's eating the page.

~sciocont wrote:

Flying creature attempting to avoid sonar would have to use the biological equivalent of RAM (radar absorbing material). Their body covering would scatter and absorb sound waves.
also, If you don't mind, could you downsize your avatar a bit. I like it but it's eating the page.

Okay, I changed it. And is there any creature on Earth that uses RAM, or are we going to have to conceptualize a way a creature could do it?

sonar = raycasting = slow

Sonar could be simplified by taking all nearby objects (including plants/trees) and making the sonar's accuracy inversely proportional to the amount of feedback received (in other words, the more or bigger the objects around are, the less accurate would it be).

As for the camo methods suggested, so far I like none of them. Alaska's raycasting method would probably require multiple raycasts per creature and thus would be slow as all the hell. Taking the average colour of the environment has it's flaws too; take a forest for example - the ground and the treetops are all green, but the middle is brown. If green would be considered a camo color, then a green bird sitting on a brown trunk would be 'invisible'.

Commander Keen wrote:

As for the camo methods suggested, so far I like none of them. Alaska's raycasting method would probably require multiple raycasts per creature and thus would be slow as all the hell. Taking the average colour of the environment has it's flaws too; take a forest for example - the ground and the treetops are all green, but the middle is brown. If green would be considered a camo color, then a green bird sitting on a brown trunk would be 'invisible'.

Maybe different heights in the forest could take different values depending on tree height. Ground and canopy would be green, while in between would be brown.

This issue is quickly getting complicated, but I will say that probably the best solution for unobserved background simulation is taking the average color of an environment. For player observed (ie, in range of the player) ray-casting may be the only realistic solution. If we keep the range low, maybe it won't take too much CPU? And aren't we going to be raycasting anyway when creatures check to see what they can see? Maybe I don't understand the concept of raycasting and alternatives, but I don't see any way around this issue other than generalizing the ecosystem and creature colors.

GhengopelALPHA wrote:

This issue is quickly getting complicated, but I will say that probably the best solution for unobserved background simulation is taking the average color of an environment. For player observed (ie, in range of the player) ray-casting may be the only realistic solution. If we keep the range low, maybe it won't take too much CPU? And aren't we going to be raycasting anyway when creatures check to see what they can see? Maybe I don't understand the concept of raycasting and alternatives, but I don't see any way around this issue other than generalizing the ecosystem and creature colors.

I like this.

Raycasting for actual gameplay, averagres or multiple averages for evolution purposes.

Quote :

Raycasting for actual gameplay, averagres or multiple averages for evolution purposes.

Sounds good.


As for the raycasting, the issue is that for the visibility test, you only need one raycast for each creature. When you check for visibility, you need to make a few more random raycasts on the body and then another few raycasts around it, and then compare the results of those. If the general color of the creature was stored, we could use it to skip all the raycasts on the body. That still leaves us with the rays casted on the background, though, a problem if we use any realistic view distances.

A solution may be randomly casting a few shortrange rays from the viewed creature itself and then comparing the result with the creature's stored average color value.

Commander Keen wrote:

Quote :

Raycasting for actual gameplay, averagres or multiple averages for evolution purposes.

<==that's kind of what I was getting at.

Sounds good.


As for the raycasting, the issue is that for the visibility test, you only need one raycast for each creature. When you check for visibility, you need to make a few more random raycasts on the body and then another few raycasts around it, and then compare the results of those. If the general color of the creature was stored, we could use it to skip all the raycasts on the body. That still leaves us with the rays casted on the background, though, a problem if we use any realistic view distances.

A solution may be randomly casting a few shortrange rays from the viewed creature itself and then comparing the result with the creature's stored average color value.

Sounds good.

Now, shapes.

Tenebrarum wrote:

Now, shapes.

No.

~sciocont wrote:

Tenebrarum wrote:

Now, shapes.

No.

Then won't sonar reliant organisms have inherently better chances?

Tenebrarum wrote:

~sciocont wrote:

Tenebrarum wrote:

Now, shapes.

No.

Then won't sonar reliant organisms have inherently better chances?

Yes, but sonar is energy-intensive, whereas having eyes uses no energy, so sonar is only useful if your eyes are not. (like in the dark and in murky waters)

~sciocont wrote:

Tenebrarum wrote:

~sciocont wrote:

Tenebrarum wrote:

Now, shapes.

No.

Then won't sonar reliant organisms have inherently better chances?

Yes, but sonar is energy-intensive, whereas having eyes uses no energy, so sonar is only useful if your eyes are not. (like in the dark and in murky waters)

Ah! And there balance is found.

And we're not going to include mimicry I assume. Too difficult.