期刊名称:ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN:2194-9042
电子版ISSN:2194-9050
出版年度:2010
卷号:XXXVIII - 1/W17
出版社:Copernicus Publications
摘要:3D city models have a wide range of applications such as mobile navigation, urban planning, telecommunication, and tourism. Thegeneration of high resolution Digital Surface Models (DSMs), in particular from airborne LIDAR data have found a major attractionof researchers in this field due to their high accuracy and high density of points. In this paper, a new approach is proposed for thegeneration of 3D models of buildings that are extracted from a high resolution DSM. Particularly the complex buildings containingseveral smaller building parts are discussed. The building parts are individually modeled by a projection-based method and mergedto form the final 3D model. The following building parts are considered to be modeled in the proposed approach: (1) flat roof, (2)gable roof, (3) hipped roof, and (4) mansard-shaped roof buildings. The complex buildings are decomposed into several partsaccording to the number of existing ridge lines. The ridge line extraction begins with image reconstruction by geodesic morphology(Arefi and Hahn, 2005) extracting the pixels of the highest part of the building segment, i.e., local maxima. Next, a Random SAmpleConsensus (RANSAC) (Hartley and Zisserman, 2004) based technique is used to fit straight lines to the ridge points. The pixels ofthe building part regarding to each ridge line are projected onto a plane which is defined based on ridge direction. According to thetype of the roof, a 2D model is fitted to the projected points. The 2D model is converted back to 3D by extruding it orthogonally tothe projection plane. After reconstructing 3D models for all building parts, they are merged to form the overall 3D model of thebuilding.
关键词:Image reconstruction; geodesic dilation; local maxima; projection based; 3D model; building roof