Abstract: The selection of Permanent Scatterer (PS) points is the key technology of Ground Based Synthetic Aperture Radar (GBSAR) deformation inversion. However, when using the traditional threshold method to select PS points, there is a problem that each region has different sensitivity to the threshold. To solve the problem of missing or wrong selection when selecting PS points, this paper proposes a model based on attention network to process radar sequence data for PS point selection, compared with the recurrent neural network (RNN) and long short term memory (LSTM) by screening PS point of three regions. The experimental results show that the real-time performance of the attention network based model is better than that of the RNN model, and its accuracy is higher than that of the LSTM model. Therefore, the model based on attention network has more advantages in PS point selection.