Automatic building detection from high resolution satellite images

An approach was developed to update the buildings of existing vector database from high resolution satellite images using image classification, Digital Elevation Models (DEM) and object extraction techniques. First, the satellite image is classified using the Maximum Likelihood Classifier (MLC). The classified output provides the shapes and the approximate locations of the buildings. Next, a normalized Digital Surface Model (nDSM) is generated by subtracting the Digital Terrain Model (DTM) from the Digital Surface Model (DSM). The differentiation of the buildings from the trees is carried out using the Normalized Difference Vegetation Index (NDVI). The classified image and nDSM are used to determine the region of interest areas to detect the buildings. The buildings in the existing vector database are then updated using the results of the above processings and the building extraction techniques. The method was implemented in a selected urban area of Ankara, Turkey using IKONOS pan-sharpened and panchromatic images. The preliminary results show that the proposed approach is satisfactory for detecting the buildings from high resolution satellite images and updating the existing vector database.