Bi-dimensional phase unwrapping is among the main critical tasks in SAR interferometry. Indeed, before the actual topography or deformation retrieval, the absolute phase values should be reconstructed from their modulo-2π wrapped version. Due to the presence of noise, the interferometric phase normally presents residues, i.e. phase jumps greater than pi on a single pixel. The residues imply that the unwrapping procedure is path-dependent, i.e. it admits different solutions. In this work, we present a preliminary investigation for the implementation of a phase unwrapping algorithm that exploits both the interferometric phase and coherence as input to a Convolutional Neural Network. The obtained results are compared with state-of-the-art algorithms.
InSAR Phase Unwrapping using Convolutional Neural Network
Scarpa G.;
2020-01-01
Abstract
Bi-dimensional phase unwrapping is among the main critical tasks in SAR interferometry. Indeed, before the actual topography or deformation retrieval, the absolute phase values should be reconstructed from their modulo-2π wrapped version. Due to the presence of noise, the interferometric phase normally presents residues, i.e. phase jumps greater than pi on a single pixel. The residues imply that the unwrapping procedure is path-dependent, i.e. it admits different solutions. In this work, we present a preliminary investigation for the implementation of a phase unwrapping algorithm that exploits both the interferometric phase and coherence as input to a Convolutional Neural Network. The obtained results are compared with state-of-the-art algorithms.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
