Abstract: PS detection is the premise of deformation monitoring using PS-DInSAR. Considering strong scattering and stability of PS pixels, this paper proposed a new PS selection method using amplitude threshold and iterative comprehensive coherent coefficient to solve the problems of existing methods such as inapplicability for small dataset. Firstly, PS candidates (PSCs) were selected using amplitude threshold. Then, the true PS pixels were selected by iterative comprehensive coherent coefficient method. The distribution, error rate and precision under different amount of images of several methods were contrasted and analyzed using 23 ERS SAR images covering Shanghai. Theoretical analysis and experimental result show that the proposed method has evident advantages for PS selection of small data set and performs better in the matter of PS distribution, error rate and rationality. The subsidence velocity error between the result of this paper and real measured data is only 1.2717 mm/a, which is precise and credible.