The study of 3‐D indoor accurate scenery modeling is an active research area. The produced model can be used in a number of virtual reality applications, in digital documentation of monuments and sites, and so on. Digital photogrammetry and CAD technology have to play a vital role in this field. Photogrammetry's contribution is mainly on data acquisition from imagery, whilst the necessary image knowledge is derived from geometry and topology of image contents. While the first is traditionally used, the second one only lately is been tackled.
In this paper, a technique is presented for modeling of indoor scenery based on digital images, photo‐derived intra‐component, geometric and topologic constraints, object‐oriented graphic databases containing 3‐D parametric models and a rough (generic) CAD model. Optionally, an absolute reference system could be applied, but for VR applications a relative reference system is adequate.
The original contribution with respect to related works in this field is mainly the introduction of Display File, Segment Table, Scene Parts Table and Constraint Table structures which deal with the constraints and “drive” a modeling control program called the Constraint Modeler. The use of these structures leads in a direct, global, portable, and semi‐automated technique for 3‐D indoor modeling. Experimental results from simulated images are presented, and the robustness of the technique is discussed.