Optimizing the exploitation of persistent scatterers in satellite radar interferometry

Abstract

Time-series synthetic aperture radar interferometry (InSAR) has evolved into a widely preferred geodetic technique for measuring topography and surface deformation of the earth. In the last decades, time-series InSAR methodologies were developed to extract information from persistent scatterers (PS) and distributed scatterers (DS). Methodologies based on DSs extract information from pixels from the natural terrain. Persistent Scatterer Interferometry (PSI) extracts information from PSs, which are found in abundance in areas with man-made infrastructure. However, a satisfactory geodetic application of these methodologies requires a complete understanding of the measurement principles, an identification of radar scatterers in the physical world, and an interpretation of the estimated deformation. Moreover, for areas not suitable for coherent imaging adding new measurements is not trivial. In consideration of the above challenges, the two main objectives of this study are: (i) to develop a systematic method to decode PSI measurements, i.e., identify PSs in the object space in order to interpret the estimated deformation (kinematics), and (ii) to assess the feasibility of encoding artificial radar scatterers, i.e. adding new measurements using radar reflectors, at places where there exists no coherent InSAR measurements. We review the contents of SAR resolution cell and the time-series processing methodologies with special focus on the Delft implementation of PSI processing. A physical interpretation of the time-series InSAR results is shown possible by decoding what the radar has measured and understanding the deformation phenomena. We employ two approaches to perform this decoding. First is to identify the source of the radar reflection by characterizing and associating PSs to a target type. By using only InSAR data, we apply an iterative classification method to discriminate radar scatterers between the ground level and elevated infrastructure. We combine the limited classification output with deformation rate and identify various deformation phenomena such as shallow compaction, no relative motion, autonomous structural motion, local land subsidence, and inter-structural deformation. In particular, we introduce a parameter known as RDI (Relative Deformation Index) to detect, quantify and analyze the regions subject to relative deformation for infrastructural stability analysis. The feasibility of this approach is successfully demonstrated with underground gas-pipe and water-pipe network monitoring applications over Amsterdam and The Hague, respectively. Second, a point-level (object or sub-object level) linking of radar reflections to real-world objects. For this step, a precise 3D position of the scatterers is derived. Applying corrections for various position error sources, accurate 3D position of scatterers is achieved for high-resolution and medium-resolution SAR imagery. A standard Gauss-Markov approach is applied to facilitate error propagation and quality assessment and control. The 2D and 3D position capabilities are validated using trihedral corner reflector field experiments. In order to precisely associate radar scatterers to physical objects, we introduce an approach to use a 3D building model of the physical objects. Linking of scatterers to parts of infrastructure is demonstrated for high-resolution and medium-resolution imagery. Finally, we propose the concept of small radar reflectors to introduce new coherent reflections. The small reflectors are designed such that they are visible from both ascending and descending imaging directions, enabling vector decomposition of deformation measurements. These small radar reflectors act as weak point scatterers. To achieve a desired SCR (Signal to Clutter Ratio), many small reflectors are distributed over an area and averaged. The detection of small reflectors is achieved by distributing them in a predefined spatial pattern. In this study, a new interferometric phase expression is derived to estimate a phase standard deviation for low-SCR and high-SCR targets. The proposed concept is experimentally validated using X-band satellite data over a grassy terrain in the Netherlands. The results indicate that distributed corner reflectors can provide deformation measurements with millimeter precision.