Camouflage

Replies to This Discussion

Permalink Reply by Ted Madden on August 19, 2015 at 8:21am

If I have a preexisting set of closed, interlocking curves does that make things easier? I get that there are a few operations that would allow me to deform the curves, but is there anything that would allow me to make them appear random so as to generate a larger group of curves that doesn't repeat?

Permalink Reply by martyn hogg on August 20, 2015 at 2:31am

One way you could investigate is by creating a random heightfield on a flat mesh and then using mesh/plane intersection to create intersection curves.

There are examples of doing this with the image sampler component... use the colour/darkness values from each pixel in an image to move points on a flat mesh in the Z direction. Then get the intersection curves of the deformed mesh with a flat plane. The example I uploaded is pointing at the image attached so you will have to save the image and point the image sampler to it by right clicking and selecting "Image"

You could use random Z values to deform the mesh instead of an image also shown in the attached example file.

If you can work out a good way of deforming the initial mesh to give you intersection curves similar to camo then this way could work. I expect it will require a more sophisticated method - perhaps using Kangaroo tools - to deform the mesh as if random spherical shapes had been pushed into the mesh rather than just randomly moving mesh verticies.

You can play with the number of points to change to smoothness of the final shapes but don't set them too large or it might take ages to compute!

Permalink Reply by martyn hogg on August 20, 2015 at 2:37am

files...

Attachments:

• lurchersmooth400.jpg, 7 KB
• Camo.gh, 70 KB

Permalink Reply by Sylvain Usai on August 20, 2015 at 5:42am

How about using perlin noise instead of random ?

Permalink Reply by Sylvain Usai on August 20, 2015 at 6:11am

should also have generated the color procedurally but meh... 

Attachments:

• 150820_Camo.gh, 21 KB

Permalink Reply by martyn hogg on August 20, 2015 at 7:22am

nice!

So I read a bit about Perlin noise. It's interesting stuff. Does it always produce this sort of undulation? i.e. you won't get neighboring points with large, opposite translations like you might get with random points?

And am i right that the Scale input is effecting the frequency of the noise?

Permalink Reply by Sylvain Usai on August 21, 2015 at 5:20pm

Hi Martyn,
My experience in perlin noise mostly comes from processing and the documentation there is quite interesting : https://processing.org/reference/noise_.html

I think this part is the key : "The actual noise structure is similar to that of an audio signal, in respect to the function's use of frequencies. Similar to the concept of harmonics in physics, Perlin noise is computed over several octaves which are added together for the final result."

So I might be wrong, but if I recall correctly, the perlin noise piles up different sinus functions (possibly in multiple dimensions) , which would explain the wave-like result.
Perlin noise can be of one, two or three dimensions (using the result of a two dimensional noise as the .z value of a point is quite handy to generate "natural look" landscapes) I guess you could also describe it as "distributed" randomness.
Now the best part about it (and that's even more true for frame based algorithm) is that if you run a perlin noise several time with the same parameters (seed, x, y, and z) value, you'll end up with the exact same result. If you are interested in the topic, you might want to dig more into "procedural generation" (application range from game design, music... Not so much architecture to my knowledge but I would love to be proved wrong). This repeatability is so interresting (to me) because that's what differentiate generative art from generative architecture (I had some quotes about this but can't recall it for memory, I could find it back though) and yet you can still achieve "random" looking results while exercing more control on it than with "real" (... Well, "real-pseudo") random.

As for the inputs, I think the time parameters sets the seed and I'm not quite sure about the scale one... (got to play more with it in grasshopper I guess)
S.

Permalink Reply by martyn hogg on August 22, 2015 at 12:59pm

Thanks for the info.

I tried applying it to this Anenome fractal tree... It alters the length of the branches and affects the direction of the branches too.

Attachments:

• 05-Anmn_tree.gh, 64 KB

Permalink Reply by Sylvain Usai on August 22, 2015 at 4:34pm

I never used Anemone, would you have any picture to show?
S.

Permalink Reply by martyn hogg on August 23, 2015 at 12:09pm

Hard to show in a picture...

Without Perlin...

With Perlin...

With Perlin and Cocoon...

Ha ha... was it worth the effort??? :)