This work presents a method to construct featured 3D terrains using information gathered from their satellites images. This data is used as keys in an image-database search. These searches use wavelets for minimize disk space and improve searching speed. In order to compose the elements with trees or bushes, for example, we use the common billboard concepts embedded with transparent technique named alpha-channel.

---

Full paper:

or

 

---

Results:

Billboard

A billboard is a simple texture-mapped geometry that increases realism when compared to the traditional intricate geometry. It rotates on the z axis in order to face the viewer all the time. During this work, many vegetation billboards prototypes were created. Some of them are shown on following:

Non-colored billboards

 

Colored billboards

 

Original	Negative	Alpha-channel
Python program uses negative for applying aplha-channel on a image

 

Terrain modeling

In order to represent in the terrain valleys and mountains, we need an hypsometric layer which gives us the entire wireframe. In addiction, a 2D texture was applied to this wireframe so that the final terrain are seen on following::

Terrain Composition

 

Classification process

The classification algorithm uses the recursive split-and-merge to find regions which are considered homogeneos. There are two stop criteria: the first one occur when the color average reaches a given threshold. The other one stops when haar-wavelet consider it homogeneos. For this stage, we use an application specially designed for this purpose.

 

Application Menu Details

 

Brazilian Itaoca regions by average method	Brazilian Itaoca regions by wavelet method

 

Brazilian Macaé regions by average method	Brazilian Macaé regions by wavelet method

Final composition

Itaoca city, Espírito Santo:

 

(a) Empty	(b) Avarege	(c) Haar-wavelet

Macaé city, Rio de Janeiro:

 

(a) Empty	(b) Avarege	(c) Haar-wavelet