Sources of 3D information are not only generated from LiDAR, nor are they purely synthesized from 2D geometries and associated attributes as in the Constructing and serving buildings 2.5D and Using ST_Extrude to extrude building footprints recipes, but they can also be created from the principles of computer vision as well. The process of calculating 3D information from the association of related keypoints between images is known as SfM.

As a computer vision concept, we can leverage SfM to generate 3D information in ways similar to how the human mind perceives the world in 3D, and further store and process that information in a PostGIS database.

A number of open source projects have matured to deal with solving SfM problems. Popular among these are Bundler, which can be found at http://phototour.cs.washington.edu/bundler/, and VisualSFM at http://ccwu.me/vsfm/. Binaries exist for multiple platforms for these tools, including versions. The nice thing about such projects is that a simple set of photos can be used to reconstruct 3D scenes.

For our purposes, we will use VisualSFM and skip the installation and configuration of this software. The reason for this is that SfM is beyond the scope of a PostGIS book to cover in detail, and we will focus on how we can use the data in PostGIS.