A Deep Learning Approach for Change Points Detection in InSAR Time Series

Interferometric SAR (InSAR) algorithms exploit synthetic aperture radar (SAR) images to estimate ground displacements, which are updated at each new satellite acquisition, over wide areas. The analysis of the resulting time series finds its application, among others, in monitoring tasks regarding seismic faults, subsidence, landslides, and urban structures, for an accurate timely response is required. Typical analyses consist identifying numerous ones that exhibit anomalous displacement, thus deserving be further investigated. In practice, this realized by selecting characterized trend changes w.r.t. historical behavior. work, we propose a deep learning approach change point detection InSAR series. designed architecture combines long short-term memory (LSTM) cells, model temporal correlation samples input series, time-gated LSTM (TGLSTM) consider sampling rate as additional information during learning. We solution lack truth developing suitable pipeline realistic data simulation. method has been developed validated through large suite experiments. Both quantitative qualitative have conducted demonstrate capabilities learned how it valid alternative statistical reference algorithm. applied real continuous project analyze Tuscany region Italy, proving effectiveness domain.