地质灾害
黄土
鉴定(生物学)
下沉
煤矿开采
地质学
采矿工程
边界(拓扑)
变形(气象学)
变形监测
地面沉降
地质灾害
地下水相关沉降
黄土高原
遥感
作者
Liuru Hu,C W Wang,Yanan Ji,Jun Deng,Chuang Song,Y F Li,Zeyu Zhang,Yu Lin,Chen Yu
出处
期刊:Remote Sensing
[Multidisciplinary Digital Publishing Institute]
日期:2026-07-13
卷期号:18 (14): 2339-2339
摘要
Mining-induced land subsidence and geohazards have become increasingly prominent in loess mining areas. Accurate identification of Active Deformation Areas (ADAs) is of great significance for mining safety, ecological protection, and geohazard prevention. The study selected the Binchang mining and used SBAS-InSAR based on Sentinel-1 SAR images from 2020 to 2025 to derive cumulative deformation and velocity. Three ADA identification approaches, including ADAfinder, MCAS-TO, and Light-UNet deep learning method, were quantitatively evaluated using IoU, F1-score, Boundary_F1-score, Precision, Recall, Area Error (%), and comparatively analyzed. ADAfinder identifies subsidence centers but produces fragmented results. MCAS-TO improves boundary continuity but is constrained by a 2% area threshold, while the deep learning method achieves the highest boundary detection accuracy and captures ADAs expansion and merging processes. Spatially, the identified ADAs are mainly distributed in the central and southern parts of the mining area and are highly consistent with underground coal mining faces. Comprehensive analysis of the three ADA identification results reveals the spatiotemporal evolution of mining-induced deformation. The results demonstrate that mining activities are the dominant driving factor controlling the formation and evolution of ADAs. The findings can provide important technical support for geohazard prevention in loess mining areas.
科研通智能强力驱动
Strongly Powered by AbleSci AI