A method for the semiautomatic detection of the topology of building floors represented as CAD drawings stored in vector file format is presented in this paper. This method involves the detection of walls and joint points amid walls and openings, and the search of intersection points amid walls. To give support to the wall detection process, this paper introduces the wall adjacency graph (WAG), a data structure created to detect walls from sets of planar segments contained in architectural floor plans. Wall adjacency graphs allow us to obtain a consistent and exhaustive set of walls very quickly (less than one second for real floor plans). A generalized version of the wall adjacency graph is also presented to deal with some of the limitations of the initial WAG. Algorithms for the detection of joint points and wall intersection points are presented as well, based on the analysis of the geometry from the input CAD drawings. Moreover, all this process works appropriately with both straight and circular segments. The obtained floor topology can later be used as input to generate 3D models of buildings, which are widely used on virtual cities, BIM systems and GIS.

This work has been partially supported by the Andalusian Government, the Spanish Ministry of Science and Innovation and the European Union (via ERDF funds) through the research projects P07-TIC-02773, TIN2007-67474-C03 and TIN2011-25259.