对流层大气延迟一直是限制合成孔径雷达干涉测量（Synthetic Aperture Radar Interferometry，InSAR）技术形变测量精度的主要因素之一。基于含有共同日期的干涉相位也包含相同的大气延迟相位分量这一事实，相位堆叠（Common Scene Stacking，CSS）方法采用叠加求和的方式来估计共有日期的大气延迟相位，并通过迭代运算提高估计精度，具有实现简单、计算效率高的优势。本文对CSS方法估算出的大气延迟相位进行空间低通滤波，并将相位解缠与CSS大气延迟相位估计进行迭代运算，改善形变估计结果。首先，基于模拟数据分析了迭代次数和时间窗口尺寸对CSS方法对流层大气延迟改正效果的影响；然后，将本文方法应用于真实SAR数据，分析了不同大小的时间窗口对改正结果的影响，并与其它方法进行对比，结果表明该方法可以获得较好的大气延迟改正结果，改正后空间相位标准差平均降低了62%，稳定点的时间序列标准差平均降低了58%。在此基础上，对目前存在的问题进行了分析和讨论。

Synthetic Aperture Radar Interferometry (InSAR) technology has the advantages of high resolution, high precision, and wide coverage, and has been widely used in the field of ground deformation monitoring. However, the tropospheric atmospheric delay has always been one of the main factors limiting the accuracy of deformation measurement of InSAR. Based on the fact that the interferograms sharing a common scene contains the same contribution of atmospheric delay from the common scene, the Common Scene Stacking (CSS) method uses the simple summation to estimate the atmospheric delay phase of the common date and improves the estimation accuracy through iterations, which has the advantages of simple implementation and high computational efficiency. The CSS method estimates the atmospheric delay phase point by point, which will introduce more noise. Based on the characteristics of the atmospheric delay phase as a low-frequency component in the spatial dimension, the estimated atmospheric delay phase by CSS is spatially low-pass filtered in this study, and the phase unwrapping and CSS atmospheric delay phase estimation are iterated to improve the deformation estimation results. Firstly, the influence of different parameters on the tropospheric atmospheric delay correction results of CSS is analyzed based on simulated data. The results show that satisfactory results can be obtained after five iterations. Then, this method is applied to real SAR data. The comparison of the results of different time windows indicates that using a larger time window will make the time series of stable points smoother, but at the same time, it will lead to distortion at both ends of the time series of deformation points. The proposed method is compared with spatio-temporal filtering, GACOS, and IPTA methods, and it is found that the correction result of the CSS method is significantly better than that of spatio-temporal filtering and GACOS. After correction by the CSS method, the spatial phase standard deviation is reduced by 62% on average, and the time-series standard deviation of the stable points is reduced by 58% on average. Compared with IPTA, the CSS method can get similar results and improve the problem of underestimation of deformation. On this basis, the applicability of the CSS method for the vertically stratified tropospheric delay is discussed. The simulation results show that the CSS method can only correct the turbulent mixing delay, but cannot effectively remove the vertically stratified delay. Therefore, the CSS method is not suitable for the steep mountainous areas where the vertically stratified component is obvious and the atmospheric delay presents a seasonal oscillation trend.