Enhancing MTInSAR Phase Unwrapping in Decorrelating Environments by Spatiotemporal Observation Optimization

Abstract

In multitemporal InSAR (MTInSAR) processing, phase unwrapping (PhU) is a vital procedure, which affects the accurate deformation retrieval, especially in environments suffering decorrelation. Although a great surge of work focuses on solving the integers of 2π. associated with the wrapped observations (i.e., the inputs of unwrapping algorithms), the quality of inputs deserves an equal attention, as it is a basis for a reliable unwrapping. In this letter, we seek to improve the PhU accuracy by enhancing the entire quality of differential phase observations. To select an optimal interferogram stack efficiently, we propose a quadtree-based coherence estimation method for fast evaluating the interferogram quality, and then develop a strategy that constructs redundant networks connecting SAR images and points, respectively. The proposed strategy utilizes the combination of minimum spanning tree (MST) and triangular closure to determine the pair of image/point included in the network. We validate the effectiveness of our method by Sentine-1 SAR data over Shenzhen airport, where decorrelated scatterers with weak backscattering energy in runways challenge a reliable subsidence extraction. The cross-comparison indicates the importance of the quality of inputs that affects the point connectivity in both temporal and spatial dimensions.