Coherent scatterer detection analysis over different test sites using TerraSAR-X images

Abstract

The identification of scatterers having a point-like scattering behaviour in SAR images has become an important topic with a wide range of applications in different scenarios, ranging from image calibration to object characterization and information extraction. In particular, this work is focused in the detection of the so-called coherent scatterers (CS’s), which are commonly associated with high scattering amplitudes, deterministic phases, high spectral correlation and high interferometric coherence values. Additionally, they are widely unaffected by multiple scattering effects and geometrical distortion, allowing a more direct interpretation. The advantage of CS’s compared with other point scatterers is that their detection is possible even on a single image basis. In the InSAR literature, a similar approach for scattering analysis and information extraction exploits the so-called permanent scatterers (PS’s), a different class of point-like scatterers. The PS’s detection is carried out by using large time series of acquisitions, and typically PS’s are required to be temporally stable. On the other hand, in the CS detection methods no assumptions are made about their temporal stability, making their usage appealing in fastly temporally variable scenarios like dynamic natural scenarios. So far, three different approaches were available for the CS detection, based on spectral correlation techniques by exploiting sublook images (i.e. images using only a portion of the available bandwidth) in range or azimuth direction. Recently, a new approach based on the generalized likelihood ratio test (GLRT) has been introduced. After an extensive theoretical performance analysis using the different detection methods and comparing the new proposed approach with the previous ones, the results have shown that the GLRT detector can achieve a better performance level than the other detectors, provided the same degradation of the spatial resolution. In this paper, the potential of the new CS detection method is tested on real data, over different scenarios (i.e. natural and urban environments), using TerraSAR-X images. In concrete, three different test sites are considered: for natural scenarios, the Helheim glacier (Greenland) and the Etna volcano (Italy) have been selected; and for urban area, the city of Paris (France) has been chosen. In particular, the performance analysis with the experimental data will be compared with the previous theoretical analysis and the impact on polarimetry will be analyzed in the test sites where polarimetric data is available.