Synthetic Aperture Radar Interferometry: Separation of atmospheric artifacts from effects due to the topography and the terrain displacements

摘要

This paper reports the new concept of possibly applying independent component analysis (ICA) on Synthetic Aperture Radar (SAR) satellite images for separating atmospheric artifacts from effects due to topography and terrain displacements. Specifically, the FastICA algorithm is applied on simulations of SAR interferograms with the purpose of extracting the different independent sources. Results show the existence of significant correlations between estimated and original components, with correlation coefficients above 0.9 and statistical confidence level above 99.9%. These findings suggest that ICA might provide a useful tool in SAR data processing, with a specific crucial usefulness in cases of an absence of ground truth knowledge, as in the cases of insufficient meteorological information at specific observational times or in satellite monitoring of remote lands. Applications on real data show that the topographical component is automatically derived by the FastICA algorithm for whatever real data set. What is different is that the extraction of terrain displacements may require some a priori information for separating different kinds of landslides and that the use of possible semi-blind ICA/FastICA approach might be considered, dependent on the specific data set.