关键词: 合成孔径雷达干涉测量, 形变监测, 时间序列InSAR, 多项式模型

Abstract:

Abstract: Time series Interferometric SAR (InSAR) techniques represented by permanent scatterers InSAR and small baseline subset approaches overcome the decorrelation limitations associated with traditional repeat-path differential SAR interferometry, thus have been gradually put into operational uses for ground deformation mapping. It is usually assumed that the deformation process can be modeled as a dominant linear component plus a nonlinear residual component when time series InSAR techniques are used. Whereas, if the real deformation scenario presents strong nonlinearity, this kind of deformation model may bring out erroneous results. This paper focuses on the deformation model of time series InSAR analysis. At first, the process of solving the interferometric phase equations and estimating the linear deformation rate is analyzed for a typical time series InSAR analysis. When the reality of deformation is deviated significantly from a linear model, and at the same time the density of extracted point targets is not good enough, the linear deformation rate can not be estimated accurately. Then, on the basis of the famous Weierstrass approximation theorem, we propose a polynomial deformation model, that is, the whole deformation will be represented by a polynomial plus the residual rather than a straight line plus a nonlinear component. Also the method to solve the interferometric phase equations under the polynomial deformation model is given. The proposed method is tested to map the ground subsidence of Taiyuan, Shanxi province of China. Totally 23 ALOS PALSAR images acquired between 2003 and 2009 are processed with the small baseline approach. In comparison, the small baseline approach with both the linear deformation model and a three-order polynomial deformation model are conducted. Both results of subsidence retrieval are compared with the leveling observation. It is demonstrated that the small baseline approach with the 3-order polynomial model can not only achieve more accurate deformation estimate, but also generate denser point targets. Since a continuous process can always be better approximated by a higher-order polynomial than a lower-order one, the proposed polynomial deformation model has the potential of replacing the wide-used linear deformation model for time series InSAR analysis.