Fusion of hyperspectral images and digital surface models for urban object extraction

The topic of this thesis is the usage of remote sensing hyperspectral images and digital surface models for urban object extraction. A method for rectilinear building polygon extraction is proposed, which accounts for edge probabilities from both datasets. The edge probabilities are detected in a linear scale space and combined by a Bayesian fusion. They are introduced as weights in the adjustment of building polygons. A new quality measure for the evaluation of the extracted building polygons, named PoLiS metric, is defined and compared to the community accepted measures.