A Deep Learning Application for Deformation Prediction from Ground-Based InSAR

Ground-based synthetic aperture radar interferometry (GB-InSAR) has the characteristics of high precision, temporal resolution, and spatial is widely used in highwall deformation monitoring. The traditional GB-InSAR real-time processing method to process whole data set or group time sequence. This type takes up a lot computer memory, low efficiency, cannot meet timeliness slope monitoring, perform prediction disaster warning forecasting. In response this problem, paper proposes series based on LSTM (long short-term memory) model. First, according early monitoring GBSAR equipment, InSAR (PS-InSAR, SBAS, etc.) obtain initial information. According calculated previous stage atmospheric environmental parameters monitored, model predict delay at next time. phase removed from interference phase, finally residual unwrapped using domain unwrapping algorithm solve deformation. predicted are added amount current moment. only needs difference map moment, which greatly saves can realize variables. reliability proposed verified by ground-based SAR Guangyuan landslide Sichuan Province.