为了更有效、更可靠地探测地表线性和非线性形变,估计SAR成像时的大气影响,本文提出了基于优化永久散射体网络的邻域差分建模方法,详细讨论了PS组网、网络优化以及估计地表形变与大气影响等参数的数学模型与计算过程,并基于Matlab环境编程实现了这一方法.最后以上海陆家嘴地区为例,选取26幅ERS-1/2卫星SAR影像进行实验,探测出实验区在1992-2002年间的线性形变速率、非线性形变量、总的区域形变场及各成像时刻的大气影响值,与地面观测结果具有较高的一致性.实验结果表明,本文提出的PS邻域差分建模方法是可行的.相应的计算程序是可靠的.

This paperproposes an approach ofmodelingwith a permanent-scatterer(PS) network (MPSN) based on the neighborhood differential idea, ami ing at effectively and reliably detecting both linear and nonlinear ground deformations and estmi ating atmospheric effects by differential synthetic aperture radar interferometry with permanent scatterers (PS- DInSAR).Firstly,the triangular irregularnetwork (TIN) ofPSs isgenerated based on thePSs mi age coordinates. Deformations are correlative both temporally and spatially, butatmospheric delays are only correlative spatially. Furthermore, the shorter the intervals ofspace and tmie series are, themore the correlationsofdeformations and atmospheric delayswill strengthen.So the phase difference ofneighboringPSs can depress the atmospheric influence. Then the neighboring differentialphase model(NDPM) is constructed based on the neighborhood ofPSs among theTIN, and the corresponding setofneighboring differential phases is generated. What are unknown in NDPM are residual elevation difference Δhijand mean velocity differenceΔvijbetween two neighboring PSsiandj.Secondly,theTIN is twice optmi ized according to distance(<1km) and multi-mi age phase coherence(γ) between two neighboringPSs,and the neighboring differentialphaseshardly influenced by atmosphere delays are selected to resolve theNDPM by using a smi ple periodogram,.i e. to estmi ate the residual elevation difference Δhijand themean velocity differenceΔvijbetween two neighboringPSs iand.j After the estmi ation of ΔhijandΔvij, the elevationΔhand themean velocityvof every PS are respectively calculated through network adjustmentbased on the optmi ized TIN.Finally, the PS sphase residueΔrescan be calculated. Then the temporaland spatialcorrelationsofdeformations and atmospheric delays are taken into accountagain, and the appropriate filters are selected to discrmi inate the followingphase from phase residueΔres,(1)PS satmospheric phases in differential interferograms, (2)PS s atmospheric phases in SAR mi ages,(3)PS sno-linear deformation phases.A setofcomputerprograms are developed inMatlab environment to mi plement the above approach.Lujazui in Shang-hai is selected as the study site,and the expermi ents are performed using 26 SAR mi ages taken byERS-1/2 from 1992 to 2002 to estmi ate linearmotion rate,nonlinearmotion magnitude, total deformation distribution, and atmospheric delay for each SAR mi age. The obtainedmeasurements are very consistentwith the ground-basedmeasurements.Based onmodelingwith PSs TIN, the neighboring differential phases hardly influenced by atmosphere delays are se-lected to resolve the NDPM. The results are more reliable and precise. The expermi ent results show that the proposed modeling andmethods are viable, and the computation software is reliable.