A Deep Learning Based Solution for Persistent Scatterers Detection

Persistent Scatterers Interferometry (PS-InSAR) has become a popular method in remote sensing because of its capability to measure terrain deformations with very high accuracy. It relies on multiple Synthetic Aperture Radar (SAR) acquisitions, to monitor points with stable proprieties over time, called Persistent Scatterers (PS)[1]. These points are unaffected by temporal decorrelation, therefore by analyzing their interferometric phase variation we can estimate the scene’s deformation rates within a given time interval. In this work, we apply two incoherent detection algorithms to identify Persistent Scatterers candidates in the city of Focșani, Romania. The first method studies the variation of targets’ intensities along the SAR acquisitions and the second method analyzes the spectral proprieties of the scatterers. The algorithms were implemented on a dataset containing 11 complex images of the region covering Buzau, Braila and Focșani cities. Images were acquired by Sentinel-1 satellite in a time span of 5 months, from October 2014 to February 2015. The processing chain follows the requirements imposed by the new C-band SAR images delivered by the Sentinel-1 satellite (launched in April 2014) imaging in Interferometric Wide (IW) mode. Considering the particularities of the TOPS (Terrain Observation with Progressive Scans in Azimuth) imaging mode[2], special requirements had to be considered for pre-processing steps. The PS detection algorithms were implemented in Gamma RS program, a software which contains various function packages dedicated to SAR images focalization, analysis and processing.

Spaceborne Multi-Temporal Synthetic Aperture Radar (SAR) Interferometry (MT-InSAR) has been a valuable tool in mapping motion phenomena in different scenarios. Recently, the capabilities of MT-InSAR for risk monitoring and preventive analysis of heritage sites have increasingly been exploited. Considering the limitations of conventional MT-InSAR techniques, in this study a two-step Tomography-based Persistent Scatterers (PS) Interferometry (Tomo-PSInSAR) approach is proposed for monitoring ground deformation and structural instabilities over the Ancient City Walls (Ming Dynasty) in Nanjing city, China. For the purpose of this study we utilized 26 Stripmap acquisitions from TerraSAR-X and TanDEM-X missions, spanning from May 2013 to February 2015. As a first step, regional-scale surface deformation rates on single PSs were derived (ranging from −40 to +5 mm/year) and used for identifying deformation hotspots as well as for the investigation of a potential correlation between urbanization and the occurrence of surface subsidence. As a second step, structural instability parameters of ancient walls (linear motion rates, non-linear motions and material thermodynamics) were estimated by an extended four-dimensional Tomo-PSInSAR model. The model applies a two-tier network strategy; that is, the detection of most reliable single PSs in the first-tier Delaunay triangulation network followed by the detection of remaining single PSs and double PSs on the second-tier local star network referring to single SPs extracted in the first-tier network. Consequently, a preliminary phase calibration relevant to the Atmospheric Phase Screen (APS) is not needed. Motion heterogeneities in the spatial domain, either caused by thermal kinetics or displacement trends, were also considered. This study underlines the potential of the proposed Tomo-PSInSAR solution for the monitoring and conservation of cultural heritage sites. The proposed approach offers a quantitative indicator to local authorities and planners for assessing potential damages as well as for the design of remediation activities.

Mine operational safety is an important aspect of maintaining the operational continuity of a mining area. In this study, we used the InSAR time series to analyze land surface changes using the ICOPS (improved combined scatterers with optimized point scatters) method. This ICOPS method combines persistent scatterers (PS) with distributed scatterers (DS) to increase surface deformation analysis’s spatial coverage and quality. One of the improvements of this study is the use of machine learning in postprocessing, based on convolutional neural networks, to increase the reliability of results. This study used data from the Sentinel-1 SAR C-band satellite during the 2016–2022 observation period at the Musan mine, North Korea. In the InSAR surface deformation time analysis, the maximum average rate of land subsidence was approximately > 15.00 cm per year, with total surface deformation of 170 cm and 70 cm for the eastern dumping area and the western dumping area, respectively. Analyzing the mechanism of land surface changes also involved evaluating the geological conditions in the Musan mining area. Our research findings show that combining machine learning and statistical methods has great potential to enhance the understanding of mine surface deformation.