Monitoring Deformation of Elevated Transport Infrastructure Using InSAR and Deep Learning

Time Series Interferometric Synthetic Aperture Radar (TS-InSAR) technique has been widely applied in earth observation field. However, there are still challenges in the fine monitoring and interpretation of long-distance, wide-span urban elevated transport infrastructures. In this study, we present a case study on the Shanghai maglev, where we mosaicked InSAR time series by jointly estimating orbital errors across two frames of TerraSAR-X images. This approach enabled us to obtain extensive, long-term InSAR observations along the track. Additionally, we used deep learning methods to cluster the observed deformation patterns. The cluster results were then interpreted with structural knowledge and environmental factors, demonstrating that deep clustering on TS-InSAR analyses can enhance monitoring efficiency and automation. This data-driven post-processing approach offers a new solution to using TS-InSAR analyses in urban infrastructure monitoring.