A novel fuzzy based method for 3D buildings modelling in urban satellite imagery

In this paper a fuzzy based method for 3d modeling of tall buildings and towers using a single satellite image is addressed. Firstly vegetations based on green layer of image and amount of gradient and image intensity were filtered out. Afterwards, buildings are segmented based on the method presented in [13]. In the next step, according to the fuzzy membership functions, obtained from structural context, the roof of each building is segmented from building walls. To make a 3d model of buildings, the side of building walls that present their height is manually selected by the human operator. Other buildings are modeled by the set angle. To benchmark the algorithm it is applied to some images from Sydney, Australia. The experimental result shows that the method could efficiently model the buildings which have at least two sides obvious in the satellite image.