Discrimination of different sources of signals in Permanent Scatterers technique by means of Independent Component Analysis

The analysis of multi-temporal SAR data-sets encountered large interest in the remote sensing community during the past few years.The main effort goes toward the extraction of ground displacements signals by means of differential interferometric techniques.In this opera- tional framework an important processing step concerns the estimation and subtraction of signals due to atmospheric artifacts and processing errors.In the present work we apply the technique of Blind Source Separation (BSS) by using the algorithm of Independent Component Analysis (ICA) to Permanent Scatterer processing in order to perform the separation of different signal components.Preliminary investigations are carried out both on simulated and real ERS-1/2 data and results are reported and commented. I. INTRODUCTION The permanent scatterers (PS) technique (1) allows the analysis of the phase information over single objects and thus typically on man-made structures, characterized by a high temporal phase stability. Where such object are present, even when surrounded by low-coherence condi- tions, a non-conventional spatio-temporal phase analysis can be performed, together with an accurate atmospheric phase screen estimation and removal. This new technique, although very promising as an operational tool for the analysis of long series of SAR images, represents still a challenge in termsof computational burden and exhibits modelling aspects not yet fully understood. The main problemsencountered when dealing with data on moun- tainous, un-urbanised areas involve the steep topography causing considerable DEM errors (∆h(x, y)), increased atmospheric effects compared to flat areas, and increased geometric distortions; also, the scarcity of man-made structures lowers considerably the number of detectable stable points, further complicating the processing (5). Many as pectsof the complicationsintroduced in the