Decomposition And Extraction Methods Of InSAR Time Series Signal Based On Independent Component Analysis

Interferometric Synthetic Aperture Radar(InSAR),a new geodetic technology developed since 1960 s,can provide observation support for monitoring the deformation of geological disaster,researching the earth dynamics and deformation mechanism with the advantages of large range,high precision and high spatial resolution.However,the InSAR observations generally include deformations caused by various geological activities/factors,in which the hidden different deformation information and processing errors are difficult to quantify and separate,which greatly hinders people from further analyzing and studying the surface movement and its dynamic process by using this technology.By using mathematical geometric model or geophysical model to describe the process of surface deformation,different deformation signals can be distinguished.However,the existing parameter estimation methods generally use a single geometric and physical model,and it is difficult to take into account the influence of model error for complex geophysical processes,and need a large number of accurate external environmental variables and a priori information,which greatly limits the accuracy of deformation separation.For this reason,this paper introduces independent component analysis(ICA)into InSAR time series analysis.ICA can make full use of the high-order statistical independence of different deformation and error signals in InSAR time series to extract various deformations and error signals caused by different geological activities.In view of this,this article studies the extraction and decomposition of InSAR time series signals based on ICA.The main innovative work includes:(1)Analyze and evaluate the applicability and effect of ICA in InSAR data,and proved that the separation effect of spatial pattern ICA is better than that of temporal pattern ICA.ICA itself is a mathematical method,when it is applied to InSAR time series,in addition to satisfying the mathematical definition,it also conforms to the physical laws of the actual deformation signal.This paper introduces and compares the effects of different interferogram formats,number of SLC images,maximum atmospheric delay and different ICA models on ICA separation.First,three deformation behaviors in InSAR time series,namely linear,periodic and mixed deformation,are simulated.Then,the relationship between the ICA separation results and the number of SLC images and the maximum atmospheric delay is obtained by simulation experiments.Finally,it is proved that the separation effect of spatial mode ICA is better than that of temporal mode ICA,and the ICA separation effect is the best in the case of sequential interferometric phase.(2)A sequential InSAR Data processing method based on ICA is proposed to separate the deformation signals caused by tectonic movement and human activities.Isolating the multi-sources deformations in the interferometric synthetic aperture radar(InSAR)data is essential for deformations’ mechanism interpretation.Existing reaseaches distinguish different deformations by combining heterogeneous geodetic observations(e.g.,InSAR and the global navigation satellite system(GNSS)),but they mainly emphasize the spatial pattern of different deformations,and ignore the deformations’ temporal characteristics.As one of the most powerful tools for blind signal separation,the ICA algorithm isolates different signals from the mixture based on their individual spatiotemporal characteristics,therefore is suitable for isolating different deformations from the InSAR data.In this paper,the performance of ICA for isolating different deformations is analyzed and validated by simulation experiments.Given that the Southern California,U.S.,has been experiencing complex deformations due to such as the tectonic and anthropogenic displacements for a long time,the ICA is used here to derive different deformations(i.e.,the tectonic and anthropogenic displacements)in the Southern California from ascending/descending Sentinel-1 InSAR LOS(line-of-sight)observations.Results indicate that both the tectonic motion with large-scale spatial pattern and linear temporal characteristic and the anthropogenic motion with concentrated spatial pattern and seasonal temporal characteristic can be successfully isolated by the ICA.(3)A temporal InSAR Technology Based on wavelet multiresolution analysis and ICA is proposed to successfully separate the orbit error from the derived tectonic deformation.The influence of orbital error is systematic,which can reduce the reliability of deformation monitoring.In this letter,we propose a method to isolate the orbital error from the multitemporal InSAR(MT-InSAR)derived tectonic deformation based on the wavelet multiresolution analysis and independent component analysis(ICA).Starting from the sequential interferometric phase of unwrapping,the tectonic deformation and orbital error are firstly extracted from the interferometric phase by wavelet analysis based on their longwavelength spatial patterns,and ICA is then used to isolate the orbital error from the tectonic deformation according to the different temporal characteristics of the two types of signals.In the simulation experiment,the root-mean-square error(RMSE)of the isolated orbital error is 2.6 mm.Experiments with real data in Southern California show that the proposed method can successfully separate the orbital error from the tectonic deformation,and the InSAR deformation rates are in good agreement with the GPS observations.