A symmetry-guided mirror reconstruction method for filling large gaps in TLS surveys of mining subsidence

Surface subsidence induced by coal mining is one of the most prominent anthropogenic geological hazards in mining areas. In terrestrial laser scanning (TLS) surveys of subsidence basins, point cloud data are often incomplete due to unavoidable surface conditions, including complex topography, steep slopes, and occlusion by man-made structures. To address this issue, a mirrored raster reconstruction (MRR) method is proposed to reconstruct missing DEM data by exploiting the inherent symmetry of subsidence basin development in near-horizontal coal seams. To improve applicability under practical engineering conditions, a systematic deviation angle (SDA) is introduced to account for deviations from ideal symmetry. The 208 working face was selected as the study area, where coordinated TLS and UAV LiDAR observations were conducted at multiple time periods. TLS- and UAV LiDAR–derived DEMs acquired on March 18 and April 14 were first differenced to obtain the corresponding subsidence basin DEM. The missing regions in the April 14 TLS-derived DEM were then reconstructed using the MRR method, and the reconstruction results were validated against synchronous UAV LiDAR observations. The comparison yielded an RMSE of 0.23 m and an RRMSE of 7.18%, demonstrating the reliability of the proposed method. The applicability of the MRR method was further evaluated using TLS data acquired on June 15 under actual engineering conditions. Based on the reconstructed DEM, key angular parameters of the subsidence basin were inverted and compared with contemporaneous GNSS measurements. After reconstruction, the angle of draw along the H-line differed from the GNSS-derived result by approximately 7°, while along the C-line, the boundary angle and angle of draw differed by approximately 10° and 5°, respectively. Compared with conventional spatial interpolation methods, such as inverse distance weighting (IDW) and Kriging, the proposed approach replaces stochastic interpolation with a deterministic reconstruction process constrained by subsidence mechanics. This method effectively mitigates TLS data gaps caused by field occlusions and provides a scalable solution for routine, large-scale subsidence monitoring in mining areas. It also offers a physically grounded framework for addressing data incompleteness in geotechnical remote sensing applications.