Güneşşel Isi Haritalamasi için 3B Nokta Bulutlarinda Çatilarin Bulunmasi

Solar energy is one of the best renewable source of energy, since it is found plenty in nature and it has the least negative impacts on the environment compared to other sources of energy. Recent initiatives encourage the adoption of solar energy by providing a set of tools to facilitate the purchase and installation of solar panels. Using findings of such initiatives for assessing a city's solar energy potential via solar mapping could make implementation of urban planning possible. Modeling urban environments can be a difficult task and accuracy of urban solar mapping depends significantly on the 3D model. However, information on roof structure is essential in order to perform accurate estimations of solar radiation. In this paper, a 3D point cloud model of an urban area is reconstructed from the 2D images collected by an unmanned air vehicle flying over the region. The 3D point clouds are segmented into meaningful areas such as roofs, building surfaces, roads and trees. The performance of the proposed algorithm is tested on a dataset collected over a small region in Istanbul Technical University campus. Markers are placed over a predetermined number of locations in the region of interest and the accuracy of these control points are improved by static GPS registration. 2D images collected during the flight are registered to reconstruct a 3D point cloud model of the region and then they are segmented using the proposed method in this paper. Experimental results indicate that roofs can be segmented with high accuracy.