Subsidence estimation from Tomographic SAR data using Deep Learning

Interferometric Synthetic Aperture Radar (InSAR) provide an effective way for identifying the deformation with a series of complex SAR images. In this paper a deep learning based approach for subsidence estimation is explored. More precisely, we explore the potential of Tomographic SAR Neural Network (TSNN), which is a fully connected neural network for height estimation, for deformation measurement. Experimental results on simulated data demonstrate that TSNN trained with the SAR image stack and the elevation is capable to detect the deformation areas and classify the deformation velocities with high accuracy.