An improved distributed scatterers extraction algorithm for monitoring tattered ground surface subsidence with DSInSAR: A case study of loess landform in Tongren county

•An approach aimed to detected reliable pixels in the loess area is proposed.•The use of time series characteristics can increase the precision of detection.•The approach can obtain landslide information more comprehensive and detailed. In order to effectively detect the detailed subsidence of tattered ground surface composed of many small fragments with the distributed scatterer interferometric synthetic aperture radar (DSInSAR) technique, a fast and accurate distributed scatterer extraction (FADSE), as an improved distributed scatterers extraction algorithm, is proposed and demonstrated in this paper. The emphasis of FADSE is on the improvement of accuracy of extracted DSs and detection efficiency as well. For the purpose, nonparametric estimation and parametric estimation methods are combined into FADSE to fast identify as many accurate statistically homogeneous pixels (SHP) as possible. Then the thresholds of homogeneous pixel number and coherence coefficient are adjusted to select DSs from SHPs. The validation of FADSE was performed in the case of loess subsidence detection in Tongren county, Qinghai Province of China, using 20 Sentinel-1A SAR images acquired between February 2016 and June 2017. Moreover, FADSE was compared with the Kolmogorov-Smirnov algorithm and Fast Statistically Homogeneous Pixel Selection method. Results show that FADSE is capable of efficiently extracting more DSs that are accurate and the detailed subsidence of tattered ground surface can be accurately detected.