相位萃取技术可最小化多基线InSAR处理中干涉像对的去相关噪声，已成为雷达干涉测量研究领域中热点技术之一。本文首先设计并实现了一套专门针对小数据集的基于相位萃取技术的多基线InSAR处理流程。整个流程分为两大模块。第一个模块为称为预处理模块，通过图像配准、地形相位移除等操作生成相位萃取技术所需的零基线SAR图像堆栈。第二个模块首先基于多视像元生成相干矩阵，并且在此基础上完成相位萃取操作。然后，利用小基线集技术的思想在萃取相位上分离出各相位稳定点上的形变情况。为了保证结果可靠性，在这一模块中还加入了大气相位预估与移除的步骤。最后，该流程被应用到了5景拍摄于太原地区的PALSAR图像上。实验结果表明，相位萃取技术在小数据集情况下仍然能够有效提升数据的相干性，这有利于多基线InSAR输出结果密度的提升。

Multi-baseline InSAR techniques have demonstrated their great potential in topographic mapping and ground surface deformation monitoring. In order to minimize the decorrelation noise between stacked SAR images in multi-baseline InSAR processes, the phase reconstruction technique has been developed recently and has become one of the hotspot techniques in radar interferometry. Due to budget limitations and unstable SAR image acquisition frequency, a lot of multi-baseline InSAR applications have to be carried out based on small image datasets. Researchers have made every endeavor to address this problem, some targeted multi-baseline InSAR processing strategies have been therefore developed. Unfortunately, there are few literatures discussing the application of phase reconstruction to small image datasets at this stage. This paper aims to evaluate the effectiveness of the phase reconstruction technique on a small SAR image datasets. A targeted multi-baseline InSAR processing scheme was designed and applied to real data. The main idea of phase reconstruction technique is to reform phase observations along a SAR stack by taking advantage of a maximum likelihood estimator which is defined on the coherence matrix estimated from each target. The proposed multi-baseline InSAR processing scheme is divided into two modules. The first one is named as “pre-processing module”, which generates the zero-baseline SAR image stack required by the phase reconstruction technique via a series of operations including image coregistration, topographic phase component removal, and so on. The second one firstly constructs coherence matrices based on multilooked pixels, thereby conducting phase reconstruction operations. Subsequently, it isolates ground surface deformation signals based on reconstructed phase observations by taking advantage of the small baseline subset technique. Noted that an atmospheric disturbances estimation and removal step was involved in this module in order to assure the reliability of the output measurements. The proposed scheme is subsequently applied to five PALSAR images acquired over Taiyuan, ShanXi Province, China. During the experimental process, the performance of the phase reconstruction technique in the case of small image subsets was analyzed in different aspects (e.g. the signal-to-noise ratio of the FFT based orbital fringe estimation process, the number of residues contained in phase unwrapping networks). The corresponding annual deformation rate field was presented, as well as the distribution of additional points obtained after the application of the phase reconstruction technique. The results has demonstrated that the phase reconstruction technique can effectively improve interferometric coherence even in the case of small image datasets, which is beneficial to the proliferation of the density of multi-baseline InSAR results. It must be noted that the size of the coherence matrix is relatively small in the case of small image datasets. Thus the precision of the phase reconstruction results is likely to be influenced by the low coherent elements of coherence matrix. In order to make the phase reconstruction technique work with small image datasets better, future works should try to eliminate the negative impacts of low-quality elements on the phase reconstruction procedure.