User localization using a typology of 3D queries

Today, the 3D data is almost exclusively used as a support for visualization purposes. In other words, 3D worlds are populated by implicit information that can be understood mostly by human actors. A human sees that the scene contains building, trees and other meaningful objects, whereas an agent accessing the scene only detects a set of polygons or other geometric primitives. We aim at giving meaning to 3D models in such a way that agents (i.e. not necessarily human actors) can visualize and make use of the information embedded in these models.