Vegetation tests

Published September 30, 2005
Advertisement
Yesterday i tried to implement quickly some trees on the landscape. Unfortunately, the test didn't work out very well.

When generating the last levels of the terrain LOD, i'm placing some trees randomly over the patch surface. The trees are made of two crossed quads with a texture on it. For each vertex of the current terrain patch, i determine the density of trees based on the altitude and slope of this vertex. The tree is then randomly displaced (to avoid seeing organized rows/columns of trees) and randomly rotated around the vertical axis.

There are three problems with that:

1. Individually, each tree is fugly - in the future i plan to procedurally generate real 3D trees with LOD, and impostors at high distances.

2. Because i haven't optimized the code, it's awfully slow.

3. The placement of the trees is not perfectly matching the textures on the ground, since the textures parameters (altitude/slope) are modified with noise values, but the trees on the CPU aren't.

I'm not going to spend more time on that now, as i just wanted to make a quick test to get an idea of the potential problems.

I'm now going to start on the integration of the terrain engine, the textures prototype, and the planets/atmosphere from space.
Previous Entry Forums opened
0 likes 2 comments

Comments

Burnhard

How will you determine the character of the terrain? Presumably this will come down from the higher level procedurally generated elements (solar systems, etc.), so for example, one planet will be covered with a particular bio-type (distribution dependant on your noise function and terrain algorithms) when another may not.

To generate unique fauna procedurally would also be exiting. There are many interesting forms that can be created in this way.
September 30, 2005 06:21 AM
Crayfish
The screenshots look really good, the first one especially looks realistic, the only fault is that the trees seem a bit too sparse. I think it will be fantastic when you put in the imposters as you definitely need to have an intermediate phase of vegetation modelling between the texels and the 3D trees you're planning to procedurally generate. I don't think the distance before using imposters need be that high either, even at 1km forests tend to look like a mass of green textured treetop "blobs".

Hoping you're finding all this feedback encouraging!
September 30, 2005 07:14 AM
You must log in to join the conversation.
Don't have a GameDev.net account? Sign up!
Profile
Author
Advertisement
Advertisement