Abstract
Automated house extraction from aerial photographs is one of the most critical steps in updating GIS databases or developing digital photogrammetric systems. In this paper, a method for extracting houses from the large-scale aerial imageries of urban areas mainly based upon both the region-based and line-based stereo matching schemes is proposed. Wavelet transform is utilized to detect edge features from images. Hough transform is employed to connect fragmentary line segments to improve reliablity in region/line-based stereo matching. Finally, houses/buildings are extracted by recognizing height difference between houses/buildings and terrain surface. The experimental results obtained with the method are demonstrated.
