Abstract: Synthetic aperture radar （SAR） plays a crucial role in the health monitoring of urban buildings. However， in two-dimensional imaging， it encounters challenges such as geometric distortion （overlap， shadowing， and foreshortening） because of the complexity of building structures. As SAR uses the geometric deformations in two-dimensional projections to obtain three-dimensional structural information， 3D imaging becomes highly significant in building health monitoring. This paper proposes a Bayesian estimation-based rotational millimeter-wave SAR 3D imaging model， as a new approach for building 3D point cloud reconstruction and high-precision height calculation. To address the image quality degradation caused by sidelobes and speckle noise， a multiangle millimeter-wave radar system was developed for data acquisition through rotation along a linear trajectory. A hierarchical matching strategy based on height hypotheses was employed to solve the Range-Doppler （R-D） equation， using correlation matching to improve efficiency. To address point matching errors， a hybrid distribution model based on stereo matching was developed. Geometric offsets were modeled as target errors， to minimize error distribution， optimize elevation， and accelerate 3D imaging. Experiments using a 77 GHz millimeter-wave radar for multiangle data acquisition and 3D point cloud imaging demonstrated an imaging accuracy of 0.412 m compared with LiDAR point cloud heights， verifying the method’s effectiveness and imaging mode reliability.