Erosion effect..

[bhaveshpandey]

Inspired by Peter Claes's Erosion effect, I wanted to achieve the same effect..

I was hoping to turn it into something like an Erode tool but initial tests made it clear that it would require a different methodology....

This has been achieved using VOP SOPs and POPs (I also tried OdForce Attrib Sim Accumulation asset but later did the entire thing with POPs and VOP SOPs)

I would love to have some feedback

thanks.

(this is the first time I feel I can share my work without any reservations..so I'm pretty happy )

Edited September 15, 2009 by bhaveshpandey

[bhaveshpandey]

i see the file's not playing till the end?? even on youtube..thats weirdly interesting

[Macha]

I'm interested in erosion too. Perhaps I could use it as my next project.

I think you need to figure out what parts of your geometry need to erode and what part doesn't. Obvious factors would be slope and soil stability.

There's a certain slope at which structures start to become unstable. If you add noise to these and weigh them with, say $BBY or something, then you might have an adequate source geometry to work from.

I'd then simply let a peak sop act on the geometry and see what happens. If that works correctly I'd start thinking about how to break of chunks in a more natural manner.

[bhaveshpandey]

Hey Macha, actually my approach was way more direct than this..Although what I did also comes with its own Limitations..

I looked at erosion as simple deformation of geo and at the same time the place getting eroded tends to lose mass..i.e it gets eaten away...(not to say that mass disappears..it just disintegrates and disperses in form of dust or smaller pieces or whatever)

So to fake this effect what I used was a VOP SOP to move the point positions of the geometry..

The good thing about this is VOPs give you a lot of control..can use multi threading..

The disadvantage of my approach was inherent to its methodology..I used the Normals to calculate and alter the point positions..so Its essential that All the geometry has all its normals facing in one direction (ie. away from the geometry or towards the geometry but not both)..for example a sphere or a Box or Torus etc..

of course the place getting affected by erosion can be decided using Colors or primitive nos etc.

cheers

Edited September 17, 2009 by bhaveshpandey

[Macha]

of course the place getting affected by erosion can be decided using Colors or primitive nos etc.

Hey, let me pick up on this throw-away comment! You wrote it just at the end almost as a bythought but in fact it is the most important thing! Erosion is not random and your effect will never work well if you think of it like that.

Erosion is beautifully procedural and that is exactly why Houdini is the perfect tool for it. The random part, the slapping on of colors and wild noisy attributes is just the icing on the cake, the tip of the melting iceberg.

Edited September 18, 2009 by Macha

[bhaveshpandey]

Hey Macha I agree with you to an extent on this one..

But this was never a physically correct model of erosion

So to fake this effect what I used was a VOP SOP to move the point positions of the geometry..

There are endless possibilities to come to the same result each with its own pros and cons..

What makes one way more viable than the other is a totally different story from person to person and requirement of the project..

Thats why the I mentioned the fact that the place getting affected by this effect could be driven by color(which could again be decided based on calculations as well as hand painted ones) i.e if you wanted more artistic control over it OR it could be driven by attributes based on calculations to get some other pattern..

I guess there's some misunderstanding (probably??)

perhaps I should have named this as "Weathering" as this effects more or less deals with disintegration and deformation of portions of a geometry..

sorry..my bad

Edited September 18, 2009 by bhaveshpandey

[pclaes]

heh, I'm away for two weeks and this pops up .

I'm honored you got inspired and I think you are off to a good start!

Things I considered when doing my effect were mainly:

- control of where the effect takes place

- control over accumulation (as in particles that go tend to slide down more were other particles have gone before) - only the sop solver seemed a real option for this... but that also makes it slower as you are now solving.

- avoiding the geometry to self-interpenetrate... try it with a different shape or let your sphere become extremely "deformed". You really do need some kind of iso-surface approach if you want to get arcs and holes (like a swiss cheese).

- if your surface is eroding, you might want to consider doing your texture in object space, so the noise does not stick to the surface (as in uv space) but you actually go for the "swimming textures" effect as you are trying not to deform but eat away the surface.

- erosion is only the first step, you also need to deposit material, unless perhaps if you are dealing with acid. My particles would "emit blue when they collide and then slowly start emitting red until they emit nothing at all" -> blue is erosion, red is deposit. Displacement inwards and outwards is easiest and probably fastest - but a bit limiting as you can't really get arcs/holes.

If I had to do it again now I would use SDF's as collision objects, a SOP solver and use purely volumes and iso-surfaces. That way you can do a cookie-operation on a volume level in vops without having to deal with polygon boolean problems. Also using custom volumetric fields you could define the areas in full 3d where to erode faster or not. And also use custom fields on the particles to guide where they should flow to.

good luck with it!

[bhaveshpandey]

thanks a lot Peter!

some really nice tips out here

Yes, since I'm displacing the points i run to a point where the geometry starts to inter-penetrate..

I'm not really familiar with the way Iso-surfaces work but I'll give it a try and update soon!

its pretty difficult for me to understand your solution using SDFs and SOP solver..but hell I will give it a try..

thanks a ton for the tips

Edited September 23, 2009 by bhaveshpandey

[bhaveshpandey]

hey Peter I'm finally able to understand a bit about your 'SDFs and iso-surfaces' as a solution..I guess I'll give it a try and update later but I'm sure this will be a better but a more time consuming process (as in SOP solvers and Iso-surfaces etc..will take down the speed..)

but yeah i guess thats one way to get accurate solves and avoid inter penetration of geometry..

thanks a ton for your help!

cheers..

Edited November 9, 2009 by bhaveshpandey