Multi-Point Isolines

[craftycurate]

It would be great to have more than 2 heights in an isoline, perhaps even to use a graph or spline editor to set the isoline shape e.g.

- the "top view" can set the X\Y (along\down) position of the points
- a side view can set the Z position (height)

And the final shape of the isoline would simply be connecting up the points using linear, spline, bezier or other curves, as in the graph editors found in many software applications.

Thanks
Richard

[cajomi]

You are right, that would be possible and a great advantage. On the other hand it would make the isolines more complex. And do not forget: The result of using an isoline is always a combination from the terrain generation algorithm (with random parts), the filters and the line.
So, a very detailed isotool would not deliver the exactness you would expect, if you have such controls.
And if Geocontrol would deliver such exactness, it would be an enormous work to create a simple mountain. Equal to sculpt it in a polygon modeller.

So at all, I am not sure, what way the isolines will take in version 2. But I found, a z-axis or profil tool would help, to create different slope defined shapes. I am thinking about a typcical "filter" tool, which would set the height of a two colour isoline correspondending to the length of the line.

[monks]

I know you said originally Johannes that you didn't want an implementation too close to WM. That's understandable, -more technical than artistic, but I think the advantages of a more 'vector-like' line tool is pretty large.
For eg, Vectors will give you more control as said by CraftyCurate for graph editing. Also if you include support for existing standards ike svg, ai, -even the GIS shp format, the possibilties really open up.
For example, shp format supports z values in the vertices. There are massive datasets out there available for download which could be incorporated into GC. Here's a link to some ideas I had recently:
http://terrain.cg-arts.org/forum/index.php?topic=89.0

We have TG2 now going to support georeffing- in the future. It will initially support the geotiff. Shp format will be great for a TG2 workflow.

Also, if you have vectors, having the abilty to draw vectors *without* defining z info would be desirable. These vectors could be used to pick up information from any dem (including real world dems) open in GC. This requires that you have a 'drop vertices to terrain' function which drapes a set of vertices to the underlying terrain, transferring the z info from the terrain to the vertices of the vectors.

Against that you have usability. Personally I really like the isolines for their responsiveness. The isolines are like painting. Vectors are like modelling- they can be a bit too precise. If you got vectors to be as responsive as isolines, that would be great- maybe beziers could be.

Also, you have to really provide some additonal support for vectors because when you model terrain via vectos with z info -ie, as networks of ridges, rivers, you really need to be able to attach new vectors to existing vectors *at the correct height*. When modelling in Wilbur's tess tool, connecting vectors at he correct z, takes about 50% of modelling time.

Overall, imo the advantages of vectors outweigh the disadavantages.

monks

[cajomi]

I disagree.

Isolines are no vectors and do not behave so. They are more a sketch, and this will stay.
Of course there is a need of more tools, like connecting, very important. But this will not make the isolines behave like vectors or real 3D modelling.
What I meant, was a simple transformation over the length of the isoline with a filter shape. So, this would not give an x/y/z relation. For this relation, you would need projections of the isoline from front/back and left/right. What I am planning is a projection over the length of the isoline, the direction of the line has no influence.


Again: the secret of a well working tool for creating fast and easy terrain like structures, is, to forgot the polygon like way. To create in a technical way a terrain is much to time intensive. I am really surprised, how much time you are willing to spent for creating a terrain, which absolutley exact follows a very unexact map.
The isolines work like a scetch, and you get great differences with different seeds for the noise. This makes it possible, to create a terrain, which equals a typical natural terrain within minutes. But it makes it nearly impossible, to create the matterhorn exact with isolines. Importing exact vector shapes of a existing landscape, what should be the sense? Simply load the DEM into the final application and you have it, or convert it and then load it into a terrain generator for further developement.

[monks]

Isolines are no vectors and do not behave so. They are more a sketch, and this will stay.

I agree. I think it would be folly to replace the isolines with vectors. I was more suggesting compliment isolines with vectors. Well it looks like you're doing this- although you're not using vector formats pre se but rather you're own implementation. No, I think what you have in mind sounds promising.


Again: the secret of a well working tool for creating fast and easy terrain like structures, is, to forgot the polygon like way. To create in a technical way a terrain is much to time intensive. I am really surprised, how much time you are willing to spent for creating a terrain, which absolutley exact follows a very unexact map.

I absolutely agree with this. Trust me, I've got damn near to trying the polygons method in Wilbur and it stinks! :) Wilbur's tool is only have finished anyway. No, this is all about water transport- this idea is aimed primarily at addressing that.

Importing exact vector shapes of a existing landscape, what should be the sense? Simply load the DEM into the final application and you have it, or convert it and then load it into a terrain generator for further developement.

I knew you were going to say this :) but the point would be you can connect the vectors up to new vectors/ existing vectors, edit the vectors, etc. In this case you could preserve correct water transport over the surface whereas if you combine dems in this way, you are never guaranteed of doing this. Correct water transport will be increasingly important in the near(ish) future- ok a couple of years maybe?- Havok is currently researching physical, interactive rivers.
I think that maybe terrain as vectors needs a completely new app to prototype it to really see if it could be viable: the full solution of a continent (or planet) described in more of a GIS language. I understand why you're reluctant to go down this road, so that's cool. One for another day. :)

monks

[cajomi]

Well, I really do not mean, it is a question of technical solutions. Of course a vector solution is possible.
And I am thinking sometimes about a possibility to implement an obj import, so that the first rough shape could be done with a polygon modeller.
But, at the end, we are more moving forward to terrains with higher resolution, up to 32000*32000 points. We more need solutions for overhangs, more information about the 3D part of a terrain.
So I expect, that all solutions, which mostly depends on heightfields with only one heightfield projection, are a dead end.
I think, with version 3 I will change this. But also then, the goal is not to give you tools, to create a overhang by hand, with polygon modelling functions, but again, with sketch like functions. GC will produce the needed overhangs, the informations in high sloped areas. To say it in other words:
To create a landscape is the work of the terrain editor, not the work of the user. To give the user the possibility to create it folllowing an idea about a landscape, that is the point.
For me, the greatest fault of the new "information" society, is, that more informations are not more informations. That is the reason, why human beings are able to filter informations. That is the power of mankind. Not the mass of informations, but the quality of filters. The idea, of "all is possible" is a dead end. More is not better. Mankind is going down very rapidly, because of this wrong idea.
Live is not improved by "Google Earth", but by taking the time to go outside, and really be a part of the earth. Google Earth give us the feeling of possesing the earth, but with using it, we loose the world, and get separeted. Wrong way.
And so, this feeling of holding the earth in your hands, by simply adding more and more informations, should never be possible with GeoControl. It is for artists, to give them the chance, to make ideas "real", the landscapes in the brains and dreams. And these landscapes are an addition to realtiy, not a replacement of realtiy.

[craftycurate]

Live is not improved by "Google Earth", but by taking the time to go outside, and really be a part of the earth. Google Earth give us the feeling of possesing the earth, but with using it, we loose the world, and get separeted. Wrong way.
And so, this feeling of holding the earth in your hands, by simply adding more and more informations, should never be possible with GeoControl. It is for artists, to give them the chance, to make ideas "real", the landscapes in the brains and dreams. And these landscapes are an addition to realtiy, not a replacement of realtiy.

This reminds me of the theory of "Hyper-reality" proposed by Jean Baudrillard - that, for example in an image-driven age that people would live by imaginary realities that were treated as more real than reality e.g. soap operas, "reality TV", and so on.

This is a constant danger for 3D art - that it represents an attempt by the artist to create a alternative reality to real life.

[cajomi]

Yes, that is what I meant.

But I believe in more realities than one. And the artist shows up other realities. And to communicate about this realities, 3D art is one of the best.

On the other hand, I have found, the more you work on landscapes, the better you watch them in real life. So, it is not the artist in danger, but the viewer.
But for me the greatest danger is the wish to possess reality, and to replace the own reality with others.

[craftycurate]

I agree completely with you, and share your experience of increased observation of real life.

I have found myself observing nature much more closely since i discovered 3D art e.g. watching how shadows behave, how water moves with the wind, how direct sunlight and ambient light work together to light a landscape, how clouds interact with sunlight. And what is more, real nature seems more real because it is so much more alive than 3D images can produce.

In a sense the "real" reality is like the Platonic ideal to which our 3D efforts can aspire.

And of course the gap between a rendered reality and the real world prompts more questions - why doesn't that rendered scene look "real"? What does it lack? What is happening with real light that my software (or my use of it) does not capture? What is the eye seeing that is not being properly emulated? So I learn more about "real" reality by noticing how the created scene falls short. By creating 3D images, I simply remind myself that I do not really understand the world fully.

This is of course a fundamental law of science too - we didn't know about cystic fibrosis till medical science discovered people who didn't metabolise mucal proteins properly. So by finding a "gap" we discover something about reality. So it is with exploring the "gap" between reality and 3D work. We discovered about ambient occlusion for example this way - how it goes dark when ambient light hits a corner - early renderers did not copy this.

And yes, I constantly have to remember to "come back" to real life after getting excited about an project.

One of my aproaches is to produce images that could be real i.e. they are imaginary places where the water, light, wind etc behave as in real life. Perhaps I do this to understand my own reality better?

[cajomi]

I am living in the "Ruhrgebiet", 5 Millionen people, one great town. No place, where you can be alone and feel the nature, no nature at all. Parks yes, but parks are not nature.
That is my reason......

[monks]

I pretty much agree with everything you guys have said here, especially: terrain modelling
increasing one's appreciation of nature. Nature has always been a real inspiration to me- and it just makes me look at real nature and think- computers...? nah, they will never do it...Even looking at the humble and lowly moorland of Britain awes me. It's not about looking at it, manipulating it, it's about inhabiting it.

The point Crafty made about discovering the correct state/functioning of things by
observing anomalies is very true. It's the same notion as a shape being defined by the
bounding area of other shapes- which leads us back into the idea of more than one reality.
Evolution is built on both success and failure.
This is why vectors interest me because we have a similar balance of 'opposition' between
raster info and vector info. I ask, why start with a raster at all? Clearly it has advantages in that it uses a homogenous division of space and what's more that's the way people have traditionally done it- so may as well work with things as they are, but I think vectors are the other side of the coin. The GIS industry is wise to that. If you define enough points, then you can interpolate betwen them Which is really how all knowledge proceeds- cf the medical analogy.
The advantage that vectors have- or topology as a description- is that it works via the space between points as opposed to raster dems which works only on the points themselves. One could look at raster dems as data, but see vectors as information. I'm sure that's why the GIS industry uses vectors in dem analysis. Therefore since interpolation is a process rather than a static data point, there is inherently more information in a vector description- or rather more efficiently expressed and I'd say more versatile (such as scalability). And that's why I think humans see terrain as maps, as networks of places because it is a more efficient way of holding information.
So I would argue that vectors are much better for describing the data when you need
efficiency- obviously. The large initial 'rough' form of the terrain. That's really what polygonal meshes are anyway; interpolated meshes of points.
Vectors has the added advantage in that it naturally describes terrain in the way humans
see it- as networks, maps, etc. Once terrain modelling passes a certain threshold of size,
then terrain moves into that space- of geography and connected places rather than slopes and relief features. This is why mountains in very large games such as mmorpgs are used as fences- to divide and channel gameplay, not as 'mountains' at all.
I see river information as being extremely important (in theory). Again I would use the above analogy: terrain = points, rivers = space between those points. They have been largely overlooked imo by terrain modelling as a source of information about *terrain*.


And I am thinking sometimes about a possibility to implement an obj import, so that the first rough shape could be done with a polygon modeller.

This is something that we had in mind for Wilbur's Tessellation tool. I think it has
potential. A good workflow in my experience for large terrains is a 'base' mesh/dem -> refinement /procedurals.

But, at the end, we are more moving forward to terrains with higher resolution, up to 32000*32000 points. We more need solutions for overhangs, more information about the 3D part of a terrain.

Yes, indeed. I'm kind of watching with interest at what ideas come along- how things shake out. I personally think that a combination of base mesh + vectors + volumetric 'patch' or whatever will be a good balance. But they all happen to be new methods to terrain modelling!
All three strike me as being very efficient ways to describe the terrain. I don't think it
will be necessary to have overhangs everywhere- only where you require them- initially as an evolutionary step towards true volumes anyway. If someone comes up with an efficient
description which potentially allows for overhangs at every point then that would be a more
elegant solution bnut it has to allow for the inherent way that modellers edit terrains- ie, not the same over the whole surface.

monks

[cajomi]

Well, what others do or what they do not, was never a really point in my decisions.
So, that GIS uses mostly vector than raster data, is absolutly no argument to me.
But let us have a look, why they use this form of data.
So, we have to go back in historie a bit, to the times, where this systems were created. There we find, that GIS data mainly were data of owners from land. The lines, where the owner changes (bad english, I know). Of course such data are stored better as vectors, the landparts are in general more or less rectangular. The system was established, and now they use it. Even if the new data do not fit to this system.
Terrains are typcical fractal data. A vector assumes, the space between is more or less linear, but fractal datas are not linear/smoothed. That is the reason, and a good reason, to work with raster date for terrains.

[monks]

I like the alt key function to get the terrain height under the mouse position. Would it be possible to use that for an iso with more than one height?
For eg, I have a terrain and I'd like to create a river. I use the subtract in the isoline layer 'Add mode' but I would need to pick both the heights at the start and the end of the isoline- to be really convenient.

Another idea for river creation is to have a function which allows you to 'drop the isoline to the underlying terrain'- however many points the isoline has (ie it would also work for isolines with many points). If you set your isoline layer for the river to 'subtract', then you could simply create a river very quickly with 'drop isoline to terrain'. This would require that the underlying terrain is always flowing downhill for the river creation to make sense. Perhaps a better system would be to have both the multi point isoline (with graph editor or whatever) and the drop to terrain.
a suggestion:

You have a multi point isoline with the ability to control the z at those points. So if you wanted a river you would sketch the z profile so that the slope was always going downhill. then you would use the 'drop to terrain' function which takes the line and finds the height of the underlying terrain at the first and last vertices/points (begin and end of isoline). It then scales the z range in the isoline to fit within these two heights. So you would get both the overall profile scaled to fit the terrain.

Or you could do it the other way: draw mulitpoint isoline. Set the endpoints to match the underlying terrain- ie 'using the alt key. These two points are then set in the z editor (or wahetever you have)- so that they are automatically transferred. All you need to do is then manually set the other points between these two OR you could simply have a 'linear/ or other interpolate' button which did it for you.- or you could havwe river profile presets...


Developing that idea, you could set up your rivers on a few layers and drop the whole layer of isolines at once to the terrain.- like 'burning in' the watercourses.



monks

[monks]

The current isoline needs to be specialised maybe for river creation because the river bed is always flat, you need sloping river beds- good for some effects tho. Maybe the z profile used for a ridge should be used for the river bed- ie inverted?

//I'm looking at river creation and so far I've been really impressed. I've not really studied it closelyyet but I've had some great results up to now...I've just had what look like very good rivers without specifying the z profile of the isolines. You do have to be careful how you set them up but after applying the thin flows erosion filter, it's promising.



monks

[cajomi]

I am still planning a special river tool.

Do you have an example of that rivers?

[monks]

I didn't save the terrain but I'll have another go and post an image.

monks

[monks]

A quick stab at it below. The terrain is a single isoline running horizontally across to form the ridge. From these are several isolines connecting to form a river network. There are 3 sizes.
It's hard to determine whether they are all flowing downhill- but I think these rivers would be good in (for eg) something like the Geomorpher we are using at ME-DEM to illustrate the river flow.
You have to be careful with the octaves (levels in graph editor)- keeping to 100 inthe middel of octavbes tended to work here. The larger rivers were ok with the lowest octaves set. To get an idea, I recommend looking at the settings in the Glaciate filters- if you set the high octaves you can get some really incised - often unnaturally so- channels.

I did have problems when I used more than one isoline for the horizonatal hillslope part- I'll try that again tomorrow.

monks

[cajomi]

I have just bought camstasia and I am planning some video tutorials.
I have found, that the sloped isolines (2 colours) really work good for a river bed, but you have to pay attention, that the surounding area is higher.