Abstract

Offset tracking based on normalized cross-correlation (NCC) of synthetic aperture radar (SAR) intensity imagery serves as a critical technique for monitoring large-scale ground deformations. However, traditional NCC of SAR intensity imagery is susceptible to isolated high-intensity points, which can induce patch-like errors and compromise the reliability of the derived deformation fields. Existing suppression methods do not differentiate between isolated high-intensity points and those constituting structural features, which are beneficial for NCC, resulting in a substantial loss of valid offset measurements concurrent with errors mitigation. Regarding this, we proposed a method for suppressing patch-like errors of SAR intensity offset tracking. The new method initially employs Z-score standardization to rapidly screen high-intensity points; subsequently, Influenced Outlierness (INFLO) structural density analysis is utilized to identify isolated high-intensity points (classified as outliers), which are then replaced by the median values of their local neighborhood prior to the NCC computation. A method for detecting patch-like errors was also designed based on the spatial characteristics of patch-like errors, defined by abrupt boundary discontinuities and high internal homogeneity. On this basis, quantitative metrics including the patch-like errors removal rate and the valid offset coverage rate were further designed to evaluate the approach’s capability in eliminating patch-like errors while retaining valid offset measurements. Comparative experiments were conducted using simulated and real SAR data. Results demonstrate that the proposed method achieves patch-like errors suppression comparable to existing methods while significantly enhancing the retention of valid offset measurements and improving overall estimation accuracy. Specifically, in the real data experiments over the Amnye Machen and Central Tianshan test areas, compared to the logarithmic weighted NCC, the proposed method increased the valid offset coverage rates by 0.272 and 0.264, and improved the comprehensive quality indices by 0.191 and 0.184, respectively. This study represents a refinement of classical deformation estimation methodologies, offering a more robust option for monitoring large-scale ground deformation.

IPC Classification

Keywords