Building height estimation using hybrid VHR optical and SAR imagery

Today’s space borne SAR systems are capable of imaging the Earth surface with a resolution of not better than 9 m (Radarsat-1 fine mode). This will change in the near future, when the 1 m very high resolution (VHR) SAR onboard the TerraSAR-X and COSMO-SkyMED satellites become fully operational. In VHR SAR data, features from individual urban structures, like buildings, can be identified in their characteristic settings in urban settlement patterns (e.g. different residential areas, inner city centers, industrial parks).

JRC’s Information Support For Effective And Rapid External Aid (ISFEREA) action is developing its expertise in urban settlement characterization as part of its activities in support of crisis management, damage assessment, territorial management in third world countries and assessment of urbanization in African mega cities. Future use of VHR SAR is expected to become an important complimentary methodology to urban characterization based on high and very high resolution optical imagery.

The scattering phenomena of buildings in VHR SAR are mainly affected by the inherent side looking configuration of the SAR sensor. Flat roofed buildings which have the front wall parallel to the azimuth direction are characterized by a rectangular layover area, a double bounce stripe, a roof scattering area and a rectangular shadow region. The extents of the different zones depend on the dimension of the building and the incidence angle of the beam. For building with a gable roof, a stripe with strong backscatter is placed in front of the building resulting from single bounce contributions from the roof. The location and extend of this feature depends on the inclination angle of the roof and the viewing configuration. For buildings with aspect angles that are not parallel to the azimuth direction the backscattering characteristics changes primarily due to the absence of the strong double bounce effect and the angular effect on the shadow area.

In this paper we present a 3D building reconstruction methodology which is based on the hybrid usage of VHR optical and SAR imagery. From the optical data the location, the extends of the building in two dimensions and the roof type is extracted. The height of the building is determined from the SAR imagery using a SAR simulator in an iterative procedure. The SAR simulator, which has been optimized in terms of memory efficiency and computational costs, is based on ray tracing taking into account coherent backscattering from single and multiple scattering effects. Using the building footprint and roof characteristics for the model initialization, an iterative simulation procedure of buildings with different heights results in a local backscattering image which is then correlated to the local scene. The iteration stops when a match criterion has been satisfied. The main advantage of the application of a simulator in combination with the calculation of a match criterion is the simultaneous consideration of all characteristics of a building in SAR for the estimation of the height value.

The efficiency of this approach is demonstrated on a test side in Dorsten, Germany using VHR airborne optical and SAR data. The results are validated with DSMs and ancillary ground truth data sets such as building height measurements and 3D building models which were acquired during several measurement campaigns using Laser and LIDAR devices.