Hi,

l would like to generate a big amount of vegetation (I used to do it on 3 ds max), but is there a way to tell houdini to generate only on the camera frustum or I must say which is visible to camera in order to enhance performances? Please find an attachment of what I used to use on max

You can easily do this using Camera Frustum SOP and Group SOP:
http://www.orbolt.com/asset/animatrix::cameraFrustum::1.00 [orbolt.com]

It generates a very light weight closed mesh (can adjust segments too). Plug this into the second input of a Group node and use the Bounding Object option which works for grouping points only.

If I understand, what he wants is to scatter some vegetation based only to what is visible to the camera to prevent houdini to do unwanted and useless calculations

I would look into packed primitives,

https://vimeo.com/102531317 [vimeo.com] << This is 1 of 2 - make sure you watch 2 as well as it covers some things wrong in 1

and personally forget about doing anything other than cutting the terrain with a camera frustum (I would not even bother except, in certain situations), Packed primitives were made to deal exactly with this kind of situation, and any plants outside of the camera viewpoint won't cause memory overhead.

I think that culling geometry, or scattering geo based upon the camera frustum, while it is fairly easy, I think will result in a lot of issues:

1. I think your plants will pop when they go on as the enter the frustum - You could create a vop that would scale them as the enter the group but then they are scaling up

2.What happens with the shadows? You could pre-bake depth maps, but again you are asking for issues.

is there any way we can delete polys which are not visible in camera, like back side of object??