Three-Dimensional Positioning of Point Scatterers Based on Radargrammetry

Abstract

This paper presents a new technique for the retrieval of 3-D point scatterer (PS) location and resolution cell configuration exploiting only the intensity of synthetic aperture radar (SAR) images. The implemented method is based on the principle of radargrammetry and makes use of Bayesian inference, wherein, directed graphs are utilized to represent dependencies of random variables and probability density functions are modeled by particle filter representations. Compared to published radargrammetric work, the newly developed algorithm optimally exploits stacks of acquisitions taken with at least three different incidence angles. Due to the large angular diversity, the method is insensitive to atmospheric propagation effects and motion of the scatterers in contrast to the coherent phase-based techniques such as persistent scatterer interferometry and SAR Tomography. Additionally, the method estimates absolute height, which is a big advantage compared to the relative estimates provided by the phase-based techniques, which moreover encounter phase unwrapping errors and temporal decorrelation. High-resolution spotlight TerraSAR-X data of Berlin central station is used as a processing example for this technique. The test case demonstrates the unambiguous absolute height estimation of PSs and resolving of complicated scattering situations (e.g., layover) in urban areas.