Multi-temporal Synthetic Aperture Radar Interferometry (MTInSAR) techniques allow detecting and monitoring millimetric displacements occurring on selected targets exhibiting coherent radar backscattering properties. Because of the bandwidth limitation of Synthetic Aperture Radar images, pulse responses are significantly affected by sidelobes that should be identified and masked out before the selection of coherent targets candidates. In fact, sidelobes often exhibit a coherent backscattering behavior and can be hence erroneously identified as persistent scatterers (PS). These artifacts can be also clearly visible in the images, in case of strong scatterers embedded in a surrounding weak clutter environment, such as the echoes generated by ships or buildings. The actual large amount of Geospatial Big Data fosters the implementation of automatic tools for sidelobe detection and cancelation.
A number of algorithms have been proposed in the scientific literature for sidelobe reduction in Single-Look-Complex images, ranging from apodization methods to target extraction techniques. In this study we deal in particular with the Spatially Variant Apodization (SVA), a nonlinear filter based on cosine-on-pedestal weighting functions. This approach is attractive because it is able – in theory – to achieve a total sidelobe cancelation, without degrading the original image resolution, and for its computational efficiency. However, as shown in recent works, SVA suffers from several drawbacks: (i) it modifies the statistics of speckle-dominated areas, (ii) a negative bias is introduced in the reflectivity images over homogeneous areas, (iii) the amplitude and phase of the original images could be distorted by the SVA filtering. In particular, it has been demonstrated that SVA filtering compromise the quality of MTInSAR outcomes. Therefore, recent works propose the use of SVA for just identifying and masking out the radar coordinates of the detected side-lobes in the PS candidates position map.
In this work we apply SVA for proper coherent target selection in COSMO-SkyMed STRIPMAP co-polarized interferometric data takes, acquired in onshore and offshore environments. In order to highlight both advantages and limitations of this technique, we analyze the performances of SVA approaches in terms of computational efficiency and robustness to focusing artifacts (responsible of non-ideal pulse responses). In particular, we test the SVA filtering both in simulation and in complex real scenarios, where sidelobes of close strong scatterers are mixed and superimposed upon each other. We show the outcomes of MTInSAR techniques before and after the application of SVA filtering.
The research is partly co-funded by the Italian Space Agency in the framework of the FAST4MAP project (“Fast & Advanced SAR Techniques for Monitoring & Alerting Processes” - ASI Contract n. 2015-020-R.0). CSK® Products© of ASI delivered by ASI under a license to use.