AUTOMATIC RECONSTRUCTION OF BUILDING OPENINGS IN 3D MODELS USING 2D ARCHITECTURE IMAGES

• Main Author: Laksmiari Saraswati, R
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/40177
• Institution: Institut Teknologi Bandung

In conducting building 3D modeling, semantic and geometric information has an important role for various applications in visualization, such as making disaster evacuation routes to building planning and maintenance. The generally available 3D models are still in the form of a LOD2 building model, which is added to the texture of the building facade photo to each 3D model polygon. The texture only provides information in plain view but does not yet have a geometric structure of windows and doors (openings) on the facade of the building. The geometry of building openings until now is generally still done manually. This study tried to develop a method of reconstruction of open geometry in building facades by utilizing Computer Vision. Computer Vision has a role in matching window and door features taken from 2D architectural drawings of building facades. After the window and door, features can be detected, the next step is to create a bounding boxes vector of windows and doors detected. The vector of windows and doors is then reconstructed into the available LOD2 building model. This thesis research hopes for developments in the automation of geometric reconstruction of building openings that can be done efficiently and effectively. And the expected results of this thesis research are making a model of the LOD3 building from the LOD2 building model and building facade architecture. Besides that, a room plan image is used to model the LOD4 building. The resulting data is vector data and attributes of each building facade.