, 2012 
  
  
04th January, 
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
MATCHING PERSISTENT SCATTERERS TO BUILDINGS 
A. Schunert*, L. Schack , U. Soergel 
2 Leibniz Universitit Hannover, Institute of Photogrammetry and Geolnformation, Nienburger Strasse 1, 30167 
Hannover, Germany - (schunert, schack, soergel)@ipi.uni-hannover.de 
Commission VII, WG 2 
KEY WORDS: SAR, Interferometer, Monitoring, Matching, Urban 
ABSTRACT: 
Persistent Scatterer Interferometry (PSI) is by now a mature technique for the estimation of surface deformation in urban areas. In 
contrast to the classical interferometry a stack of interferograms is used to minimize the influence of atmospheric disturbances and to 
select a set of temporarily stable radar targets, the so called Persistent Scatterers (PS). As a result the deformation time series and the 
height for all identified PS are obtained with high accuracy. The achievable PS density depends thereby on the characteristics of the 
scene at hand and on the spatial resolution of the used SAR data. This means especially that the location of PS cannot be chosen by 
the operator and consequently deformation processes of interest may be spatially undersampled and not retrievable from the data. In 
case of the newly available high resolution SAR data, offering a ground resolution around one metre, the sampling is potentially 
dense enough to enable a monitoring of single buildings. However, the number of PS to be found on a single building highly 
depends on its orientation to the viewing direction of the sensor, its facade and roof structure, and also the surrounding buildings. It 
is thus of major importance to assess the PS density for the buildings in a scene for real world monitoring scenarios. Besides that it is 
interesting from a scientific point of view to investigate the factors influencing the PS density. In this work, we fuse building outlines 
(i.e. 2D GIS data) with a geocoded PS point cloud, which consists mainly in estimating and removing a shift between both datasets. 
After alignment of both datasets, the PS are assigned to buildings, which is in turn used to determine the PS density per building. 
The resulting map is a helpful tool to investigate the factors influencing PS density at buildings. 
1. INTRODUCTION (Germany). The main prerequisite is an assignment of PS to 
buildings. For that the buildings are represented by their 
In the last years Persistent Scatterer Interferometry (PSI) outlines and matched to the PS set (i.e. a residual shift between 
attracted a lot of attention as a tool for accurately mapping building outlines and the PS set is removed). The PS are then 
deformation on a sparse grid of temporally stable radar targets attributed to the closest building if their individual distance is 
(Ferretti et al, 2000), (Hooper, 2006). Especially the possibility ^ below a threshold. Since no polyhedral 3D city model of the test 
to completely or partially replace expensive and time site is available, the volume of every building is calculated 
consuming measurement campaigns like levelling is very based on a prismatic model. That is, the outline is extruded to 
appealing. In general the PS density is very high in urban areas, the mean height of the building. The resulting map is finally 
which is especially true if high resolution data featuring a used to select three interesting sites showing some factors, 
ground resolution around one meter (for instance acquired by which have a strong influence on the PS density. 
the TerraSAR-X satellite in High Resolution Spotlight mode) is 
used. According to (Gernhardt et al, 2010) densities of up to 2. DATA 
100,000 PS per square kilometre can be achieved, which makes 
even the monitoring of single buildings conceivable. However, In order to determine the PS density per building, the PS point 
a sufficient sampling for all buildings is by no means cloud is fused with map data constituting the building outlines. 
guaranteed. While some buildings accommodate a plethora of ^ In the following both datasets are briefly introduced. 
PS, others host just few or even no points. In order to apply PSI 
operationally for the surveillance of urban infrastructure, it is 2.1 Persistent Scatterer 
important to determine how good a structure under 
investigation can be monitored with the available data. Froma The PS results are based on processing of a stack of 20 
scientific point of view it is very interesting to investigate the TerraSAR-X High Resolution Spotlight images. The applied 
circumstances leading to PS. For that purpose a map indicating method follows the ideas presented in (Ferretti et al. 2000) and 
the PS density per building is very helpful to identify (Liu et al., 2009). A pixel of the SAR image is chosen as a PS if 
conspicuous cases like buildings hosting unexpectedly many or its phase is temporally coherent and if its amplitude is a local 
few PS. A quite simple but still expressive measure is the maximum. The former is a quite common criterion to enforce 
number of PS per volume. In this work we aim to map this temporal stability, while the latter prevents the selection of 
quantity for a test site located in the inner city area of Berlin several pixels per PS. As a result a set of temporally stable radar 
  
* Corresponding author. 
79