Multi-Temporal InSAR Parallel Processing for Sentinel-1 Large-Scale Surface Deformation Mapping

Abstract

Interferometric synthetic aperture radar (InSAR) has achieved great success in various geodetic applications, and its potential for ground deformation measurements on the large scale has attracted increasingly more attention in recent years. The increasing number of synthetic aperture radar (SAR) satellite systems have steadily provided a large amount of SAR data. Among these systems, the Sentinel-1 mission can be considered a milestone in the development of InSAR techniques, offering new opportunities to monitor global surface deformation with high precision, due to its wide coverage, short revisit time, and free access. However, conventional InSAR techniques have encountered great challenges in large-scale InSAR processing over wide areas because of the large computational burden and complexity. In this work, we present a novel parallel computing-based coherent scatterer InSAR (P-CSInSAR) method for automatic and efficient generation of deformation results from Sentinel-1 raw data. To achieve high parallelization performance for the overall InSAR processing chain, parallelization strategies at different levels have been adopted in the P-CSInSAR method, which has been fully addressed in this work. To evaluate the efficiency and accuracy of the proposed method, P-CSInSAR has been tested on the North China Plain regions with three adjacent frames of Sentinel-1 images, and the deformation results have been validated by GPS measurements. The experimental results confirm the effectiveness of the proposed parallel computing-based P-CSInSAR method. The proposed method can also play an important role in exploiting Sentinel-1 InSAR big data for disaster prevention and reduction.