A Novel Two-Step PSInSAR Algorithm Using No Predefined Deformation Model

Persistent Scatterer Interferometry (PSInSAR) suffers from estimation error and bias when the real displacement of PS candidates does not fit to the predefined model, e.g., linear deformat ion model. In this paper, we aim to develop a new PSInSAR scheme not applying any predefined displacement model. Firstly, the short-temporal-baseline interferogram stack is generated for residual DEM reconstruction under the assumption that the displacement between successive acquisition times can be neglected. The estimation of residual DEM is achieved through the optimization of stack coherence and the PS points are reselected using maximum stack coherence criterion. Then, after removing the topographic phase is removed from the single-master differential interferogram stack, the remainder unwrapped phases are considered as only containing displacement phase, atmospheric phase and other noise terms. In order to separate the displacement phase term from the unwrapped phase series, Empirical Mode Decomposition (EMD) algorithm is used because it is totally data-driven. This paper also gives simulat ion and real data experiments to prove the advantages of the proposed algorithm.