Extracción automática de edificios a partir de modelos digitales de superficiesel sistema Building

Resum

In this work we present an approximation to the complicated problem of the semiautomatic location, extraction and vectorial representation of buildings. This approximation is based on dense models analysis of surface and of the oriented aerial images of a determinate area. The proposed method uses as input information the digital surface models (obtained from photogrammetric process or LIDAR mapping data capture according to a very high density grid) and a previous geometric model of the form of the existent buildings in the zone. The surface models will be processed using mathematical morphology operators in order to classify ground points and non-ground points. Furthermore, the regions that have been classified as non- ground regions will be processed using invariant geometric moments in order to obtain the representative rectangle of localized object. Once the possible buildings have been detected, and their representative rectangles have been defined. The results will be refined using the complementary information provided by aerial photogrammetric imagery. Using digital image analysis procedures it is possible to extract the main borders presented in the images and applying the Hough’s transform the principal rectangle will be obtained. This rectangle combined with the heights obtained from the DSM in the same location, constitute the rectangular base prism that represent to the building. All these operations have been implemented in a own developed system, whose preliminary results are presented in this works in order to establish the capabilities and problems of the methodology.