Land-cover classification-based Persistent Scatterers identification for peri-urban applications

We illustrate, through a sample application to a difficult landslide test site, the use of a novel method to detect potentially stable objects in Persistent Scatterers SAR Interferometry (PSI). Conventional PSI processing involves selecting first-guess potential stable objects, called PS Candidates (PSC), through thresholding of the amplitude dispersion index. This method can lead, in applications to scenes characterized by scarce urbanization, to very low PSC numbers, insufficient for a successful subsequent phase analysis if their spatial distribution is very sparse. Our classification-based approach relies on the proven fact that urban areas are more likely to contain PS pixels than any other land-cover class. Therefore, using pixels belonging to the urban land-cover class as PSC is a convenient way of increasing the number of initial fiducial points while keeping false alarm probabilities to reasonable levels. Results show that PSC belonging to the urban class, selected through simple external classification algorithms, lead to more consistent results for the final PS, both in terms of spatial density, and of reliability of displacement series.