Abstract: China possesses abundant coal resources,whose exploitation has made significant contributions to economic
development. However,large-scale mining has also caused severe ecological damage and surface subsidence. Interferometric
Synthetic Aperture Radar (InSAR) has become an essential technique for monitoring surface deformation in mining areas. Nevertheless,
Differential InSAR (D-InSAR) and Offset Tracking (OT) primarily retrieve line-of-sight (LOS) and azimuthal displacements,
which cannot fully capture the true three-dimensional (3D) surface deformation. To address this limitation,this paper
proposes a 3D deformation extraction method that integrates D-InSAR with an improved offset tracking technique. By establishing
a mathematical relationship between mining-induced horizontal displacement and ground tilt,the method reconstructs
the 3D surface deformation field in mining areas. Firstly,a dynamic-window offset tracking scheme with local feature selection is adopted,in which GLCM-based texture measures,Canny edge features,and Hu invariant moments are used to extract local
image characteristics. A dynamic adjustment factor based on the coefficient of variation,together with a comprehensive feature
score,is then applied to adaptively refine the initial matching windows. Secondly,a Gaussian Mixture Model (GMM) is employed
to dynamically determine the threshold range for fusing deformation estimates from D-InSAR and OT. Finally,a proportional
relationship model between horizontal displacement and subsidence gradient is constructed based on the probability integral
method. Combined with the spatial relationship between LOS deformation and true ground displacement,this enables quasi-
3D reconstruction of mining-induced surface deformation fields from the fused LOS measurements. By processing two Sentinel-
1 images over the study area,the 3D surface deformation field was successfully retrieved. The reconstructed 3D deformation derived
from the proposed fusion method shows good agreement with the measured data,with root-mean-square errors of 0. 376 m,
0. 502 m,and 0. 430 m in the vertical,east-west,and north-south directions,respectively. Compared with 3D inversion based
solely on D-InSAR,the proposed method improves the accuracy of vertical subsidence estimation and enhances the spatial continuity
in regions with large displacements and low coherence. Moreover,it maintains consistent deformation trends and meterlevel
accuracy under both long and short temporal baseline conditions. The study results can provide methodological reference
and technical support for surface deformation monitoring and subsidence evolution analysis in mining regions.

Key words: mining subsidence,deformation monitoring in mining area,D-InSAR,offset tracking,3D deformation