Modelling Real World Objects Combining Photogrammetric Reconstruction and Semantic Description

A system approach for modelling real world objects is presented which combines the knowledge of different disciplines as photogrammetry, computer graphics, and computer vision to generate three-dimensional models without expert knowledge from as few images as possible. In addition to the common approach representing reconstructed models only by nodes, edge, and faces, this paper introduces a semantic description which allows to represent the model hierachically. This system was applied to historical monuments (Königshalle Lorsch) as well as to industrial machines in order to reuse their geometry in factory planning systems.