Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XXXVIII-4/W19, 193-197, 2011
© Author(s) 2011. This work is distributed under
the Creative Commons Attribution 3.0 License.
07 Sep 2012
O. Melnikova1 and F. Prandi2 1Czech Technical University, Zikova 4, 16 6 36, Prague 6, Czech Republic
2Fondazione Graphitech, Alla Cascata 56/C, 38123, Povo-Trento, Italy
Keywords: Buildings Detection, Rooftop Hypothesis, Image Matching, Features Extraction Abstract. This paper introduces a semi-automatic method for buildings extraction through multiple-view aerial image analysis. The advantage of the used semi-automatic approach is that it allows processing of each building individually finding the parameters of buildings features extraction more precisely for each area. On the early stage the presented technique uses an extraction of line segments that is done only inside of areas specified manually. The rooftop hypothesis is used further to determine a subset of quadrangles, which could form building roofs from a set of extracted lines and corners obtained on the previous stage. After collecting of all potential roof shapes in all images overlaps, the epipolar geometry is applied to find matching between images. This allows to make an accurate selection of building roofs removing false-positive ones and to identify their global 3D coordinates given camera internal parameters and coordinates. The last step of the image matching is based on geometrical constraints in contrast to traditional correlation. The correlation is applied only in some highly restricted areas in order to find coordinates more precisely, in such a way significantly reducing processing time of the algorithm. The algorithm has been tested on a set of Milan's aerial images and shows highly accurate results.
Conference paper (PDF, 3544 KB)

Citation: Melnikova, O. and Prandi, F.: 3D BUILDINGS EXTRACTION FROM AERIAL IMAGES, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XXXVIII-4/W19, 193-197,, 2011.

BibTeX EndNote Reference Manager XML