Abstract: The identification of Permanent Scatters (PS) is a key step in the Ground-Based Interferometric Synthetic Aperture Radar (GB-InSAR) system for acquiring deformation data, as their accuracy and quantity directly affect the final accuracy and reliability of deformation measurements. To address the issue of insufficient quantity and quality of PS selected by conventional methods in densely vegetated areas, this paper proposes a PS selection method based on multi-dimensional features. The method first calculates seven feature values for each pixel, including amplitude deviation, temporal coherence coefficient, spatial coherence coefficient, and phase stability. Then, the variance of these feature values is used as a weighting coefficient to construct an evaluation index. Bayesian estimation is employed to adjust the weighting coefficients, thereby improving the applicability and accuracy of the evaluation index. Finally, the evaluation index is used as a clustering sample to achieve PS selection. The performance of the proposed method is experimentally analyzed using radar monitoring data from Wanzhou District, Chongqing. The results demonstrate that this method enables high-quality PS selection and enhances the accuracy of phase unwrapping and atmospheric phase compensation.