Terrain Generation

The Problem

Publicly available digital elevation models (DEM) of terrain do not have very high resolution - in some parts of the world you cannot find any data at all. Your need might be to generate a model of a film location, model a golf course for a video game, or allow your soldier to see a model of the terrain that exists over the next hill so he can increase his situational awareness prior to any action.

You also do not have the luxury of time for people to create the models by using reference imagery.

The Solution

2d3, since our core competancy stems from highly accurate image registration and structure from motion knowledge, has developed a unique capability that allows a user to use nothing more than DV or HD video footage to reconstruct and texture a 3d model of the terrain being recorded. It is a single camera solution that allows for accurate models to be created in hours, not days or weeks.

We use structure from motion techniques to identify and track thousands of feature points on the terrain that are used to reconstruct a 'mesh' or model of the land. Since the same points used to create the terrain are present in the frames of video, we can also use them to register parts of the imagery to the underlying model. We can ortho-rectify a texutre and apply it to the mesh with normals facing up, resulting in a very accurately registered texture.