Abstract: To address the challenges of 3D deformation inversion caused by the limited viewing angle of single radars and the asynchronous acquisition of multi-radar systems under fire conditions, this paper proposes a multi-radar 3D deformation field inversion method integrating spatiotemporal alignment and spatial geometric constraints. Firstly, high-precision Line-of-Sight (LOS) displacements are obtained using Frequency Modulated Continuous Wave (FMCW) and differential interferometry techniques. Secondly, a common spatiotemporal reference axis is constructed. The cubic spline interpolation algorithm is employed to perform temporal resampling and synchronization correction on multi-source asynchronous displacement sequences. Subsequently, a physically consistent 3D displacement inversion model is established using the Weighted Least Squares (WLS) method. A fire-induced collapse test of a two-story steel frame with a 1:2 scaling ratio was conducted jointly with Tongji University, and the 3D inversion results were compared with monitoring data from wire displacement meters. The results indicate that the proposed method effectively eliminates systematic errors introduced by multi-radar asynchrony, and the inverted 3D displacement field is consistent in trend with the wire displacement meter data. This method achieves high-precision, non-contact monitoring of structural 3D deformation under fire conditions and is suitable for deformation early warning in complex environments.