Prediction method of surface subsidence induced by block caving method based on UAV oblique photogrammetry

Unmanned Aerial Vehicle (UAV) oblique photogrammetry has been extensively employed in mining, albeit predominantly for reconstructing three-dimensional scenes and detecting changes within mining sites, lacking predictive capabilities. Leveraging 3D real scene model data, this study presents a two-stage prediction model, merging the probabilistic integral method with recurrent neural network (PIMF-RNN), to mitigate the impact of internal and external factors on surface subsidence, thereby enhancing predictive accuracy. Building upon this framework, a methodology was developed to forecast the maximum surface subsidence height and affected area under the block caving method, offering crucial data support for mitigating hazards associated with this mining technique. Analysis of surface data from Pulang copper mine during 2018–2020 demonstrates a prediction accuracy of 91.47% for maximum surface subsidence height and 87.52% for subsidence area. This research expands the potential applications of UAV oblique photogrammetry techniques within mining contexts. Furthermore, it establishes a cost-effective and efficient operational procedure for predicting mine surface subsidence.