Research On Key Processing Technology Of InSAR For High Precision DEM

Synthetic Aperture Radar Interferometry(InSAR)has been widely used in high-precision digital elevation model(DEM)inversion and large-scale surface deformation monitoring.InSAR technology has the advantages of all-weather,all-day and high efficiency.However,the accuracy of its data processing steps will directly affect the accuracy and reliability of the DEM.Interferometric phase filtering and phase unwrapping are the two key steps in the InSAR interferometric data processing process,and the quality of these two steps will directly affect the accuracy of the DEM.Therefore,these two steps have always been the hotspot and difficulty of InSAR technology data processing research.In this paper,we study the interferometric phase filtering method for dense fringe regions,and the key problems such as the phase unwrapping method with high noise level and large terrain fluctuation.The main contents include the influence of spatial and frequency domains on the interference phase filtering results,the influence of unscented Kalman filter phase unwrapping model optimization on single baseline phase unwrapping(SB-PU)results,and some improved techniques of multi-baseline phase unwrapping(MB-PU)method model.The main research results obtained are as follows:(1)A more detailed and in-depth study of the high-precision DEM inversion method.A detailed theoretical introduction to each step in the InSAR data processing link is provided.The main sources of error in the data processing are analyzed,which provides theoretical and research basis for the study of interferometric phase filtering and phase unwrapping.(2)The interferogram filtering methods of several spatial and frequency domains have been studied in detail.Theoretical and experimental results show that most of the spatial domain filtering methods are relatively efficient,but the filtering results will have loss of phase and resolution,especially in the dense fringe regions.Frequency domain interferometric phase filtering can obtain better filtering results than spatial domain filtering,but it is still difficult to achieve filtering and avoid phase loss in dense fringe regions.Moreover,the traditional frequency domain filtering method has relatively complicated parameter settings and low filtering efficiency.These factors limit the application of this method.This paper proposes a two-step interferogram filtering method based on the advantages and disadvantages of spatial domain and frequency filtering.This method combines the advantages of spatial domain and frequency domain,which not only improves the precision of filtering,but also improves the filtering efficiency.This method analyzes the filtering window-setting problem and proposes a simple and effective filtering window adaptive method.Experiments show that the proposed two-step filtering method and adaptive filtering window setting can eliminate residuals more effectively while avoiding phase loss.Simulation and real data experiments show that the proposed method can obtain better filtering results.(3)By analyzing several theoretical models of Kalman filter phase unwrapping method,an adaptive unscented Kalman filter phase unwrapping(AUKFMPU)method combined with median filtering is proposed.Although adaptive unscented Kalman filter(AUKF)has been well applied in other fields,it has been applied for the first time in phase unwrapping.Moreover,combined with the median filter to apply to the first domestic C-band SAR satellite data processing,namely: Gao-Fen 3 SAR(GF3-SAR).The simulation and real data show that the proposed AUKFMPU method can adaptively process the model according to different pixel error.This method increases the noise robustness of the unwrapped model and improves the accuracy of acquiring DEM.Moreover,it applied to the data processing of GF3-SAR.The experimental comparison shows that the proposed AUKFMPU method has better noise robustness and obtains higher precision DEM results than other methods.(4)A MB-PU UKFPU method is proposed for the large-scale fluctuation of the terrain,which solves the problem of poor precision of DEM inversion in mountain regions.Compared with(SB-PU)phase unwrapping,MB-PU is not limited by the continuity assumption,so MB-PU can solve the phase discontinuity caused by excessive terrain fluctuation.However,most MB-PU methods are limited by factors such as poor noise immunity and low computational efficiency.Aiming at this problem,this paper applies the UKF model to the MB-PU method based on the two-stage programming algorithm(TSPA)2-D MB-PU method,and proposes a MB-PU UKFPU method based on two-step programming(TSPA-UKFPU).Specifically,the method first estimates the distance and azimuth gradients using the first step of TSPA.Then,the estimated gradients are slightly filtered by median filtering.Finally,the maximum heap strategy and the UKF model are combined to unwrap the interferometric phase.The two sets of simulation data and the real data of ALOS-L and TanDEM-X are processed,and the results are evaluated by high-resolution lidar point(ICESat)and Shuttle Radar Topography Mission(SRTM).The results are compared with the statistical cost stream unwrapping method(SNAPHU),the traditional TSPA method in SB-PU,the experimental results show that the TSPA-UKFPU proposed in this paper is a better MB-PU method.