Phase-Based Similarly Decorrelated Pixel Selection and Phase-Linking in InSAR Using Circular Statistics

Shuyi Yao,Timo Balz
DOI: https://doi.org/10.1109/tgrs.2024.3422171
IF: 8.2
2024-07-13
IEEE Transactions on Geoscience and Remote Sensing
Abstract:Circular statistics is the mathematical theory for dealing with variables distributed on a circle. The interferometric phase of distributed targets can be modeled with circular statistics due to its wrapped and pseudorandom nature. In this study, we introduce a novel adaptive neighborhood selection (ANS) method and a novel phase-linking (PL) method for distributed scatterer (DS) interferometry, both based on circular statistics principles. The proposed ANS method enables the direct selection of pixels with similar SAR interferometry (InSAR) decorrelation behaviors, called similarly decorrelated pixels (SDP), from the interferometric phases. The proposed PL method: 1) shows significant resistance to the potential departure from the fully developed speckle assumption in the SAR observations (also known as non-Gaussianity) compared to methods that rely on this assumption; 2) does not introduce substantial implementation complexity, computational cost, or numerical solution challenges compared to methods that elaborately model the non-Gaussianity, e.g., through a product model; and 3) can achieve higher consistent wrapped phase estimation precision compared to other methods solely based on interferometric phases. In addition to validating the proposed methods through simulation experiments, we also found that a combination of the two proposed methods can produce interferograms with minimal noise in a real data experiment.
engineering, electrical & electronic,imaging science & photographic technology,remote sensing,geochemistry & geophysics
What problem does this paper attempt to address?