| The frequent human underground activities and the natural geological activities, easily result in the underground fluid movement and basement empty, and then cause regional surface subsidence for the surface crust uneven stress. In the urban and densely populated area, the ground subsidence makes series of security problems, such as, the ground deformation, the building crack or collapse and the road fracture. Furthermore, it threatens people’s life and property security. Therefore it is necessary to monitor surface deformation effectively and accurately, which will make great contribution for emergency management and early warning. Furthermore, it is urgently to actively develop novel effective space observation technology and related system.Differential synthetic aperture radar interferometry (DInSAR), with some prominent advantages such as high efficiency, all-time, pantoscopic view and high accuracy, is widely applied for monitoring and investigating land subsidence field. However, to solve the negative affections including atmospheric delays, orbit errors, DEM errors and system thermal noise, Berardino has proposed the small baseline subsets method (SBAS) in 2002, for extracting the sequential deformation, it has a higher accuracy when extracting long-term ground deformation. Nowadays, it has become the hot topic in this field.However, it is the situation that sequential images are sampled sparsely cannot be avoid. Then, the rank-deficient may happen during the calculation processing of observation equation. As a result, the singular value decomposition method used for retrieving deformation may have an unreliable result. To extend the time span and keep the higher accuracy of deformation monitoring, it is necessary to implement synergistic analysis based on multi-platform imagery.To solve these problems, this paper proposes a method using the multi-platform SAR images, called model-constrained small baseline subset (MC-SBAS) method, so as to extend the time span and keep the higher accuracy of deformation monitoring. This method can improve the accuracy of deformation monitoring by adding nonlinear model. The details are as follows:(1) For solving rank deficient equations, this paper proposes the model-constrained small baseline subset method. This method can connect the deformation between independent subsets, and improve the continuity of the sequential deformation, through adding nonlinear models at each image point in the rank deficient equations.(2) Based on the characteristics of complex surface deformation, the nonlinear model consists of a linear, nonlinear, periodic, elevation error and constant composition. We try to make the experiment using 23 PALSAR images in TianJin as source data. The result shows that the MC-SBAS method and the nonlinear model are effective, and it also can ensure the robustness of the sequential deformation monitoring in the sequential image sampling sparsely or discontinuously.(3) In order to extend the time span of deformation monitoring, we establish a nonlinear model of constraint equations by multi-platform images. Simulate platform of differential phase subsets by the parameters of real SAR platform, and add 4 groups of random noise into the simulated phase. The statistical result of difference between SBAS and MC-SBAS methods and the standard value, shows that MC-SBAS error distribution in -5 mm to 5 mm range increases by 19% and the RMSE error decreases ±12.5 mm. Whereafter, the difference of the result between MC-SBAS and SBAS have been carried out. And result show that MC-SBAS method has higher reliability and precision than SBAS in extracting the sequential deformation.(4) We monitor ground deformation by ERS (acquired between 2002 and 2005, descending) and ASAR SAR (acquired between 2004 and 2008, descending) images in Southern California. Analyzing the result of SBAS and MC-SBAS, we find that they have similar deformation distribution in spatial dimensions. But in temporal dimensions, compared with SBAS (11.7 mm error), the MC-SBAS has a higher accuracy, which the RMSE error between the MC-SBAS and GPS results is 8.7cm.Theoretical and experimental studies have shown that MC-SBAS method can solve rank deficient equations and combine multi-platform SAR images to build models. Compared with SBAS method, MC-SBAS method can improve resolving precision and reliability, and also can extend the time span of deformation monitoring. |
PDF Full Text Request