Objectives Ground-based synthetic aperture radar (SAR) time series analysis mainly selects coherent pixels with strong scattering characteristics and stable phase from SAR images for subsequent time series inversion. Coherent pixels between different interferograms are variable, which makes time series analysis complicated. Therefore, it is crucial to select a high number of coherent pixels that are of high-quality to ensure precision in ground-based SAR deformation monitoring.
Methods To improve the reliability of time series analysis in ground-based SAR, this paper proposed the enhanced nonlocal coherent pixels selection (E-NonPS) method to select effective coherent pixels from the interferogram network, including full-rank coherent pixels and partially coherent pixels. Full-rank coherent pixels are selected by the nonlocal full-rank method, and partially coherent pixels are selected by setting the signal-to-noise ratio (SNR) threshold for pixels with coherence occurrences of more than N/2 in single look complex images.
Results We used E-NonPS method to test two ground-based SAR datasets to verify the effectiveness in different monitoring environments (e.g., high and low coherence areas), and compared the selection results of coherent pixels with three traditional methods, including amplitude dispersion index (ADI) method, coherence coefficient (Coh) method and nonlocal full-rank coherent pixel selection (NonPS) method. The experimental results show that E-NonPS method can significantly increase the number of selected coherent pixels, and effectively improve the accuracy, especially for low coherence areas.
Conclusions The propoesed E-NonPS method can effectively address the demands for selecting coherent pixels and ensure sufficient coherent pixels in time series analysis.
DownLoad: