Deep Learning Based Persistent Scatterers Detection: First Results

We consider the topic of multitemporal or multipass Differential Synthetic Aperture Radar Interferometry (DInSAR) for monitoring ground deformation of the built environment. A fundamental step in the processing is represented by the stage of separation of pixels corresponding to monitored Persistent Scatterers (PS) from noise pixels. PS detection is typically implemented with classical detection methods, e.g. the Generalized Likelihood Ratio Test. In this work we specifically discuss the results of first experiments for the use of Deep Learning in the PS detection.