PhD Positions at GRACE for International Doctoral Candidates

Innovative Approaches for Improved Geocoding and Localization of InSAR Deformation Signals in Infrastructure Monitoring

SAR interferometry is widely used to study earth surface dynamics and deformation associated with geodynamic processes. At the same time, it offers considerable potential for monitoring of structural behavior and deformation of critical infrastructure, including dams, bridges, and transportation corridors.

High-resolution SAR missions allow for detailed observations of individual structures. However, their limited availability and associated costs often prevent continuous long-term monitoring over wide-areas. In contrast, freely available satellite missions such as Sentinel-1 provide regular and near-global radar observations over long periods of time. Their spatial resolution is however limited, which can make the interpretation of measurements challenging, particularly for complex structures and unfavorable imaging geometries. In addition, the satellite acquisition geometry of Sentinel-1 limits the capability to spatially differentiate radar backscatter originating from within the same resolution cell. As a result, it is often difficult to determine which part of a structure a measured deformation signal represents, or whether the signal originates from the surrounding terrain.

The aim of this research project is to develop new methods for better identifying and interpreting the contributions of different backscattering objects within a radar resolution cell and for improving the spatial localization of persistent scatterers in interferometric analysis. The overall goal is to increase the reliability and interpretability of deformation measurements derived from medium resolution radar interferometry, particularly for the monitoring of infrastructure and urban areas.

While the methods to be developed should generally be applicable to a variety of infrastructure objects, the “real-world” evaluation can be done using the Enguri Dam in Georgia as a case study. A comprehensive multi-sensor dataset is available, including Sentinel-1 data, medium- and high-resolution TerraSAR-X satellite data, high-resolution ground-based SAR (GB-SAR) measurements and GNSS measurements as well as detailed 3D models. The dam provides a large and well-suited test site, combining a sufficiently complex radar imaging geometry with a well-characterized deformation signal.

The research will focus on two main parts:

Topic A: Innovative approaches for improving the geocoding of infrastructure in SAR-imagery

• Improving the geocoding of measurement points from medium-resolution SAR systems such as Sentinel-1 by simulating radar backscatter based on detailed 3D models of the observed structure.

• Investigating transfer-learning approaches to adapt geocoding between high- and medium-resolution radar systems as well as between ground-based and satellite-based SAR measurements.

• Exploring the potential of different radar polarizations to improve the identification and localization of radar scatterers.

Topic B: Innovative approaches for improving the interpretation of deformation signals in SAR interferometry

• Using tomographic approaches to spatially separate different signal contributions within interferometric time series.

• Modelling and analyzing deformation signals affected by multipath radar reflections by means of machine-learning methods.

• Integrating multi-sensor observations to estimate the full three-dimensional displacement vector of monitored structures.

The project combines satellite and ground-based radar remote sensing, 3D modelling, radar interferometry, and machine-learning approaches. It therefore provides the opportunity to contribute to the development of advanced remote-sensing methods for reliable, large-scale monitoring of critical infrastructure.

The selected candidate will be part of the Remote Sensing and Active Sensors Group at the Institute of Photogrammetry and Remote Sensing (IPF) at KIT.

Specific profile of the candidate:

• Above-average Master’s degree in a relevant field, such as geodesy, geophysics, geoinformatics or computer science

• Strong academic background in signal processing, radar imaging or remote sensing

• Proficiency in image processing and machine learning

• Interest in working with Radar instruments and data

• Excellent teamwork skills