Reduction of Building Facade Model Complexity using Computer Vision

3D Building facade reconstruction enables the implementation of delicate actions related to restoration, support and presentation of architectural structures. In particular, model complexity can be significantly reduced through the incorporation of computer vision and image processing tools, aiming towards the estimation of structural 3D feature characteristics instead of their actual GPS specification. In this paper we present a strategy for constructing reduced-complexity 3D models, which combines stereo vision images and morphological processing methods. Initially, the building skeleton is extracted and modeled with the use of Hough transformation and Active Contours. Subsequently, the building's facade depth map is derived using the stereo camera layout, for the refinement of inner skeleton details, such as the protrusion of windows or doors. These image-driven features are fused with Georeferenced data in order to complete the full building's skeleton model. The proposed reconstruction framework is tested on the site of Neoria buildings, a historical (Venetian) site in Chania, Crete, Greece.