Entity Embeddings in Deep Learning for the Detection of Anomalous Insar Deformation Signals

Maral Bayaraa, Cristian Rossi, Alfredo Kalaitzis, Brian Sheil

Abstract: A novel methodology for detecting anomalous deformation behaviour from satellite-Synthetic Aperture Radar Interferometry (InSAR) is proposed. The representation of InSAR metadata as embeddings within a deep learning framework (EE-DL) is investigated for modelling the spatio-temporal deformation response. To evaluate the performance of the EE-DL approach in SAR interferometry, we conduct experiments over a mining test site (Cadia, Australia) which has been subject to a tailings storage facility failure. This study demonstrates that EE-DL can detect and predict the fine spatial movement patterns that eventually resulted in the failure. We also compare the results with deformation predictions from a common baseline model, Random Forest.