Image-based 3D Data Capture in Urban Scenarios

Presuming that airborne imagery is available at a sufficient overlap, state-of-the-art multi-stereo matching can generate DSM raster representations at an accuracy and resolution which corresponds to the ground sampling distance (GSD) of the original images. For such secenarios recent matching software exploits the resulting redundancy and derives surface representations at a remarkable accuracy and reliability. Typically, DSM rasters are generated as a standard result at a grid size corresponding to the average pixel footprint by a rather simple fusion of the 3D point clouds from multi-view matching. While such 2.5D models are suitable for a number of applications, high resolution data capture in complex urban environments requires the reconstruction and representation of 3D representations. This is especially true while aiming at the geometric reconstruction of objects with distinct structure like urban furniture or building façades. After a brief introduction in the state-of-the-art on high density image matching for DSM computation, this generation of filtered point clouds and 3D meshes within our multi-view reconstruction pipeline is discussed for both imagery aerial cameras and camera based mobile mapping systems. The results are especially beneficial while aiming at high quality visualisations and geometric data capture in urban scenarios.