A sequential Monte Carlo framework for noise filtering in InSAR time series

This article proposes an alternative filtering technique to improve interferometric synthetic aperture radar (InSAR) time series by reducing residual noise while retaining the ground deformation signal. To this end, for the first time, a data-driven approach is introduced, which is based on Takens's method within the sequential Monte Carlo framework, allowing for a model-free approach to filter noisy data. Both a Kalman-based filter and a particle filter (PF) are applied within this framework to investigate their impact on retrieving the signals. More specifically, PF and particle smoother [PaSm; to avoid confusion with persistent scatterers (PSs)] are tested for their ability to deal with non-Gaussian noise. A synthetic test based on simulated InSAR time series, as well as a real test, is designed to investigate the capability of the proposed approach compared with the spatiotemporal filtering of InSAR time series. Results indicate that PFs and more specifically PaSm perform better than other applied methods, as indicated by reduced errors in both tests. Two other variants of PF and adaptive unscented Kalman filter (AUKF) are presented and are found to be able to perform similar to PaSm but with reduced computation time. This article suggests that PFs tested here could be applied in InSAR processing chains.