Abstract
Persistent scatterer interferometry (PSI) has the ability to acquire submeter-scale digital elevation model (DEM) and millimeter-scale deformation. A limitation to the application of PSI is that only single persistent scatterers (SPSs) are detected, and pixels with multiple dominant scatterers from different sources are discarded in PSI processing. Synthetic aperture radar (SAR) tomography is a promising technique capable of resolving layovers. In this paper, new approaches based on a novel two-tier network aimed at robust and efficient detection of persistent scatterers (PSs) are presented. The calibration of atmospheric phase screen (APS) and the detection of PSs can be jointly implemented in the novel two-tier network. A residue-to-signal ratio (RSR) estimator is proposed to evaluate whether the APS is effectively calibrated and to select reliable PSs with accurate estimation. In the first-tier network, a Delaunay triangulation network is constructed for APS calibration and SPS detection. RSR thresholding is used to adjust the first-tier network by discarding arcs and SPS candidates (SPSCs) with inaccurate estimation, yielding more than one main network in the first-tier network. After network adjustment, we attempt to establish reliable SPS arcs to connect the main isolated networks, and the expanded largest connected network is then formed with more manual structure information subtracted. Furthermore, rather than the weighted least square (WLS) estimator, a network decomposition WLS (ND-WLS) estimator is proposed to accelerate the retrieval of absolute parameters from the expanded largest connected network, which is quite useful for large network inversion. In the second-tier network, the remaining SPSs and all the double PSs (DPSs) are detected and estimated with reference to the expanded largest connected network. Compared with traditional two-tier network-based methods, more PSs can be robustly and efficiently detected by the proposed new approaches. Experiments on interferometric high resolution TerraSAR-X SAR images are given to demonstrate the merits of the new approaches.
Highlights
Persistent scatterer interferometry (PSI), an advanced remote sensing technique, has been widely applied in long-term deformation monitoring [1,2,3,4,5] in urban areas using time-coherent pixels
The major difference between PSI and Synthetic aperture radar (SAR) tomography is the number of scatterers that are assumed within a resolution cell
As the largest connected network is expanded with more manual structure information subtracted in the first-tier network, more reliable single persistent scatterers (SPSs) and double persistent scatterer (PS) (DPSs) can be robustly detected by the novel two-tier network
Summary
Persistent scatterer interferometry (PSI), an advanced remote sensing technique, has been widely applied in long-term deformation monitoring [1,2,3,4,5] in urban areas using time-coherent pixels. For the detection of PSs, one approach is based on multi-interferogram complex coherence (MICC), which uses phase-matching between the measured signal at different acquisitions and the estimated ideal response, as manifested in [17,18]. The superior performances are measured as an increase in the detection probability for a fixed false alarm level as well as an increase in the accuracy of parameter estimation Both approaches are based on the assumption that the atmospheric phase screen (APS) has been separated from the acquisitions, at least on a small scale, before the detection of PSs. real acquisitions do not satisfy this assumption. There is a huge computation burden if we directly apply WLS estimator for retrieving the absolute parameters in the whole area, especially for large network inversion In this context, new approaches based on a novel two-tier network are proposed in this paper.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
