Abstract
Synthetic aperture radar tomography (TomoSAR) is typically used to retrieve elevation, deformation, and other key information by separating scattering points of the same slant range in multiple baseline SAR images. ‘Scatter coherence’ (SC), which refers to the coherent relationship between scattering points in cells of the same resolution (e.g. the layover pixel), is always ignored in traditional TomoSAR. This neglect sometimes leads to unexpected inversions, e.g. the mixture of the signal subspace and noise subspace. In this paper, the electromagnetic expressions of a single bounce off the ground and the wall are first analysed to demonstrate the SC. An effective SC suppression method is proposed based on virtual baseline transformation (VBT) and spatial smoothing (SS) technologies. The SS method is used for smoothing across the baseline direction and for maintaining the separation between the signal and the noise. The VBT method meets the requirements of a uniform baseline of SS using a pre-whitening process. To verify the suppression effect, six Radarsat-2 Quad Fine images are used. The experimental results show that the side lobe of the spectrum is suppressed and that false alarms are excluded in the proposed method. The experiments of the ‘non-coherence problem’ in the cross-polarization channel and dense urban environment demonstrate the effectiveness of this method. In the 4D-TomoSAR experiment, the SC suppression also helps optimize the spectrum.
Acknowledgement
The authors would like to thank reviewers for their instructive comments.