[A.J]

lately, i saw a demo reel about a terrain camera mapping,
and when i see the wireframe, its exactly the same as the real 2D mountain/rocks/ other terrain object

is there any technique to do this other than traditional modelling?

[A.J]

Bump!
U guys dont understand what i mean?

[nickmarshallvfx]

There are different ways that this can be done, i'll quickly outline a few:

1) Lidar scan - A full 3D scan of a real environment can be done that will match a photograph taken from the same location pretty much perfectly.

2) Use a height map of the real area to generate a reasonably accurate 3d model and then match the camera angle from photo to 3d geo.

3) Get the geo close with traditional modelling and then do some matte painting on the photos to make them fit the geo more perfectly.

4) This isn't widely used yet, but there are image based modelling techniques that can create a pretty accurate 3d model based on an object that has been photographed from multiple different angles. This can work well for small objects and props, but I'm not sure it would apply well to large environments...

Can you share the demoreel you saw? That might help to recognise the techniques that was used.

Cheers,
Nick

[A.J]

here it is

http://www.youtube.com/watch?v=4fu7EcJFuOo
this
http://www.youtube.com/watch?v=hp9qO...eature=related

the modelling are damn accurate if u ask me

and another one with some kind of zeppelin flying through mountains, maybe u've seen this


and nick, im only familiar with your 3rd option lol
untill now, i've been using traditional modelling (bevel extrude etc.) and paint deformation
but the result are nothing close to the demoreel there

[nickmarshallvfx]

I'd say you just need to refine your models a little, its possible to get almost 100% spot-on with the models if you have the time to spend, but when it comes down it it, its often faster (and therefore cheaper) to get the 3d models close and then just tweak the photos to match the geo. Nothing wrong with that, happens all the time in production. In the examples you posted, it looks like that has just been poly modelled in the same way you are doing it already.

The method of using Lidar data is popular for large budget productions because it gives hugely detailed geo that should be totally accurate to the location (of course in reality an environment TD normally has to spend a bit of time cleaning the Lidar geo up). We used this method recently to recreate a mountain range not too dissimilar to the ones in your videos and found that with good photo ref and accurate Lidar data we could get a decent camera projected shot in just a couple of hours. Its very expensive to do though, well beyond what a hobbyist / student would want to do, and to be honest, well beyond what's needed for the shots to work in a lot of cases!

In case it helps, when you are doing the modelling, try and get the volume of the shape fairly accurate with as few polygons as possible and then add detail (or tesselate the geo etc) to areas that need refining. This will mean that when you get down to having really heavy geo, you shouldn't need to move them around much, it should just be tweaking to give surface and edge detail.

Hope it helps!
N

[A.J]

i see,
but , what technique u are referring to?
im afraid im missing something (standard bevel extrude inset taper), paint deformation, what else?