Answers are grounded in this paper and cite their sources.
Surface subsidence induced by coal mining often exhibits spatially extension. Precise identification and temporal tracking of subsidence field boundaries are essential for safe mining and damage prevention. In traditional Interferometric Synthetic Aperture Radar (InSAR) techniques, the deformation field cannot be completely segmented due to invalid deformation in the center of the mining center caused by phase unwrapping errors and large gradient deformation. Moreover, the spatiotemporal tracking of mining-induced deformations is seldom studied. To address these challenges, we propose an automated spatiotemporal tracking framework with an enhanced real-time object segmentation model, i.e., You Only Look Once version 11 (YOLOv11), based on SAR interferograms, where the backbone network is then enhanced by incorporating a Double Branch-Simple Attention Module (DB-SimAM) attention mechanism, aiming to improve feature extraction capabilities and reduce background interference. After segmentation for each SAR interferogram, time series deformation boundaries are generated to sequentially reveal the dynamic subsidence evolution. The experimental results from the coal mining region of northern Shaanxi Province with Sentinel-1A SAR images acquired from August 22, 2018, to April 7, 2019 indicate that the new method can achieve better segmentation accuracy than common models. Moreover, it can track the changes of the deformation boundary and deformation area dynamically. This study highlights the deep learning capabilities in terms of automation, efficiency, and accuracy in identifying subsidence boundaries caused by underground coal mining activities.
No takes yet. Share an insight, caveat, or question.
Lou et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: