Quantile-based homogeneity selection and similarity-weighted robust coherence matrix for distributed scatterer interferometry in mining deformation monitoring

Abstract

Coal mining can lead to a series of geological issues, making it crucial to monitor surface deformation for early hazard warning in mining areas. Distributed Scatterer Interferometry (DSI) enables the monitoring of deformation in overlying strata above mining working faces. However, the accuracy of this technology is often affected by spatiotemporal decorrelation, necessitating phase recovery processing. The Phase Linking (PL) algorithm can effectively recover the interferometric phase, but it is limited by the correctness of statistical homogeneous pixel selection (SHPs) and the utilization of spatial dimensional information, making it difficult to obtain a sample complex coherence matrix (SCCM) that closely resembles the true situation. This severely impacts the accuracy of the PL estimation results. To address this, this paper proposes two solutions: (1) For SHPs, the medcouple (MC) method is used as a robust measure of skewed distributions, adjusting the boxplot to optimize homogeneous samples in the intensity map, removing outliers and enhancing sample robustness; (2) For SCCM construction, a similarity measure between reference pixels and homogeneous pixels is used as a weight, and a weighted sum with constraint processing is applied to obtain a robust coherence matrix. Simulation and real-world results show that the proposed method can recover phase information severely affected by noise, maintain phase stability, and significantly increase the monitoring point density; its high reliability has been validated through leveling data verification.