Phase-Preserving Super-Resolution of Complex Radar Data via ADI-Guided Multi-Branch Fusion
Abstract
Ground-based radar (GBR) has attracted notable attention for monitoring structural health and geological hazards because it provides robust and precise deformation measurements reaching millimeter accuracy. However, factors such as antenna aperture size and manufacturing costs impose inherent limitations on radar image resolution. Radar super-resolution addresses these challenges by enhancing image resolution and extracting finer details from observed scenes. Current state-of-the-art methods for ground radar super-resolution primarily rely on single-branch single-channel convolutional architectures that only process the radar amplitude. However, this disqualifies their applicability to real-world scenarios, especially for deformation monitoring, where estimating the phase observations is essential. To overcome this limitation, we propose a novel approach using an amplitude dispersion index (ADI)-guided multi-branch feature fusion deep learning (DL) network. In our framework, the real and imaginary components of the complex radar data, along with the ADI, are processed through separate branches to extract complementary feature representations. These are subsequently fused with features from a main branch that processes the 2-channel complex radar input. The ADI is a coherence-aware prior, guiding the network to focus on spatially consistent backscattered signals while suppressing incoherent clutter and noise. The extracted features from each branch are adaptively fused through cascaded convolutional blocks with skip connections and dense mappings. The proposed model is applied to a real outdoor dataset, collected using the recently developed multiple-input multiple-output (MIMO) radar, NanoRadar, with a center frequency of 78 GHz. The experimental results show that our proposed model outperforms the benchmark methods for estimating the high-resolution amplitude and phase maps based on relevant evaluation criteria. In addition, we apply a deformation monitoring framework for the estimated radar images. The deformation map and time series show promising performance compared to the ground truth data.
Keywords
How to cite
Helmy, I., Schenk, A., Ismail, M., Hexel, R., & Tuxworth, G. (2025). Phase-Preserving Super-Resolution of Complex Radar Data via ADI-Guided Multibranch Fusion. IEEE Transactions on Geoscience and Remote Sensing, 63, 1–20. https://doi.org/10.1109/tgrs.2025.3622923