Potential of Using Phase Correlation in Distributed Scatterer InSAR Applied to Built Scenarios
Publication in refereed journal


摘要The improved spatial resolution of Synthetic Aperture Radar (SAR) images from newly launched sensors has promoted a more frequent use of distributed scatterer (DS) interferometry (DSI) in urban monitoring, pursuing sufficient and detailed measurements. However, the commonly used statistical methods for homogeneous pixel clustering by exploring amplitude information are firstly, computationally intensive; furthermore, their necessity when applied to high-coherent built scenarios is little discussed in the literature. This paper explores the potential of using phase information for the detection of homogeneous pixels on built surfaces. We propose a simple phase-correlated pixel (PCP) clustering and introduce a coherence-weighted phase link (WPL), i.e., PCPWPL, to pursue a faster processing of interferogram phase denoising. Rather than relying on the statistical tests of amplitude characteristics, we exploit vector correlation in the complex domain to identify PCPs with similar phase observations, thus, avoiding the intensive hypothesis test. A coherence-weighted phase linking is applied for DS phase reconstruction. The estimation of geophysical parameters, e.g., deformation, is completed using an integrated network of persistent scatterers (PS) and DS. Efficiency of the proposed method is fairly illustrated by both synthetic and real data experiments. Pros and cons of the proposed PCPWPL were analyzed with the comparison to a conventional amplitude-based strategy using an X-band CosmoSkyMed dataset. It is demonstrated that the use of phase correlation is sufficient for DS monitoring in built scenarios, with equivalent measurement quantity and cheaper computational cost.
著者Guoqiang Shi, Peifeng Ma, Hui Lin, Bo Huang, Bowen Zhang, Yuzhou Liu
期刊名稱Remote Sensing
關鍵詞distributed scatterers, built scenarios, phase correlation, deformation estimation
Web of Science 學科類別Remote Sensing;Remote Sensing

上次更新時間 2021-10-01 於 00:08