Semi-automatic selection of optimum image pairs based on the interferometric coherence for time series SAR interferometry

作者：Hongan Wu, Yonghong Zhang, Yonghui Kang, Zhong Lu, Xia Cheng · 发表于：Remote Sensing Letters · 年份：2019 · DOI：10.1080/2150704x.2019.1650981 · 被引用次数：9 · 研究领域：Synthetic Aperture Radar (SAR) Applications and Techniques、Soil Moisture and Remote Sensing、Advanced SAR Imaging Techniques

Time series synthetic aperture radar (SAR) interferometry (TS-InSAR) has been widely used for monitoring ground deformation. The selection of interferometric image pairs directly affects the quality of the interferograms and the accuracy of deformation monitoring. The Small BAseline Subset (SBAS) method has been widely adopted to form interferograms in TS-InSAR. However, the selection of low-quality interferograms or missed selection of good interferograms due to temporal and/or spatial baselines exceeding the thresholds sometimes occurs in the traditional SBAS method. This letter proposes a new method, namely, the semi-automatic selection of optimum image pairs (SASOIP), based on the interferometric coherence. By quantitatively evaluating the coherence of point targets in a small feature region, the SASOIP results in high-quality interferometric image pairs selected from all possible interferometric combinations for any two SAR images. The SASOIP method can effectively reduce the probability of poor selection and missed selection of quality image pairs. Twenty-nine COSMO-SkyMed SAR images acquired from 27 August 2011, to 29 July 2015, over Tianjin, China, are used to test the proposed algorithm. Validation with 102 levelling data points demonstrates that the accuracy of ground deformation derived from the interferograms by the SASOIP method is 23.13% higher than that based on the SBAS method.