GBSAR Geocoding Based on Bayes Theorem

Ground-based synthetic aperture radar (GB-SAR) is pivotal for high-precision deformation monitoring in infrastructure (e.g., dams, bridges) and geological hazards (e.g., landslides). However, near-field 2D-to-3D geocoding challenges persist due to geometric distortions and multi-candidate ambiguities. Traditional orthogonal projection methods, reliant on far-field assumptions, fail in complex scenarios. This study proposes a Bayesian maximum a posteriori framework integrated with Doppler projection and multi-source fusion. By mapping 3D models to synthetic aperture radar coordinates via Doppler projection and incorporating physical priors (e.g., radar incidence angles), the method constructs a joint probabilistic model to screen infeasible scatterers. Experiments achieved sub-millimeter accuracy (root mean square [RMS] <0.8 mm) in deformation mapping and identified structural vibrations (10–25 hertz) with <2% deviation from accelerometer data. The framework mitigates layover and foreshortening artifacts, outperforming conventional methods, and offers robust solutions for infrastructure safety and hazard early warning.