Drone-Based Ultra-Wideband SAR Interferometry for the Generation of Digital Elevation Models

Abstract

Across-track synthetic aperture radar (SAR) interferometry (InSAR) is a remote sensing technique that enables the generation of digital elevation models (DEMs) by combining two complex SAR images acquired with a certain across-track separation. DEMs are crucial for monitoring the Earth surface and have traditionally been obtained using space- and airborne SAR systems. However, these systems are subject to stringent bandwidth regulations, constrained revisit times, and high deployment and operational costs. Drone-borne InSAR offers a cost-effective solution for the surveillance of local areas, delivering unprecedented accuracy and resolution by leveraging wider bandwidths. As a result, it has emerged as an attractive complementary technology, offering rapid deployment and flexible revisit intervals. Despite InSAR's well-established theoretical foundation, it often relies on assumptions such as the use of narrowband signals and long distances to the targets that may not hold in drone-based applications. In addition, the low flight altitudes and stability issues inherent to drones pose significant challenges for InSAR and for DEM generation in particular, especially when SAR images are acquired from antennas located on separate platforms or from the same platform at different times. This thesis demonstrates how to generate accurate, high-resolution DEMs through multiple acquisitions performed using a drone-borne radar system with a ultra-wide bandwidth. It presents the necessary adaptations of conventional InSAR concepts to suit the specific characteristics of drone systems. In particular, the expressions for the baseline decorrelation and the critical baseline are generalized to account for wide bandwidths and large baselines, where the spectral shrinkage becomes significant. The new formulation enables improved performance after spectral filtering and aligns well with simulation results both for small and large bandwidths and baseline configurations. Design considerations and limitations of drone-based InSAR systems are discussed in detail, alongside a comprehensive performance analysis. A SAR processing scheme based on the omega-k algorithm is proposed, enabling fast processing of drone-acquired raw SAR data while maintaining a focusing quality comparable to that of the back-projection algorithm, which is the commonly applied approach. An InSAR algorithm tailored to the specifics of drone-based systems is also proposed. It exploits the wide bandwidth to support phase unwrapping using radargrammetry, accounts for the highly non-linear acquisition trajectories, and mitigates baseline errors by minimizing the height differences between the overlapping parts of DEMs of adjacent areas. Experimental results over both flat and hilly terrain validate the proposed concepts and confirm the predicted performance. The results show that DEMs with decimeter-level height accuracy can be achieved at an independent posting of 25 cm x 25 cm, using a highly cost-effective system. The work performed in this thesis lays the groundwork for a new generation of high-quality DEMs for various local-scale applications. Furthermore, it represents a significant step toward realizing single-pass, distributed drone-based InSAR systems. These systems will enable the demonstration of wideband, multi-frequency and distributed SAR concepts and applications in preparation of future spaceborne SAR missions.