Aiming at the ground-based synthetic aperture radar (SAR) deformation measurement, the conventional permanent scatterer(PS) selection method isdifficultto meetthedeformation measurementrequirementsintermsofthe quantity and quality of PS selection in the time-incoherent and complex scenarios. The article proposes a PS selection method based on bi-directional long short-term memory-convolutional neural network ( BiLSTM-CNN) , which uses amplitude dispersion and amplitude tojointly selectpositive and negative samples to constructthe training dataset, and takes the interfering phase, amplitude divergence and correlation coefficientas the temporalfeatures ofthe dataset, and then learns the PS globaltemporalfeatures and the PSlocaltemporalfeaturesby using the BiLSTM and the multi-scale CNN, respectively,andthentheglobaland localtemporalfeaturesareweighted andfused tolearnbythemulti-head self- attention ( MHSA) , and finally feature probability mapping is carried out in order to construct the PS classification model. The performance ofthe proposed selection method is experimentally analyzed by using the radarmonitoring data ofJiudaoquan inWanzhou District, ChongqingMunicipality, and theresultsshow thatthemethod improvesthenetwork accuracy,F1 score, recall, precision, and other indexes, and improves the quantity and quality of radar image PS selection.