Related Experiment Video
Updated: May 28, 2026
09:19
Measuring the Structure, Composition, and Change of Underwater Environments with Large-area Imaging
Published on: April 18, 2025
MAPUNet: Multi-scale attention for InSAR phase unwrapping in mining areas
Baojing Zhang1, Wenfu Yang1,2, Xiaowei Zhang1
1Shanxi Coal Geological Investigation and Research Institute CO., LTD, Taiyuan, China.
Plos One
|May 26, 2026
Summary
A new multi-scale phase unwrapping (PU) method, MAPUNet, accurately monitors large mining area deformations. This advanced technique improves surface deformation monitoring, offering a new reference for precise geological surveys.
Area of Science:
- Geodesy
- Remote Sensing
- Geophysics
Background:
- Mining activities cause significant deformation gradients, leading to incoherence in radar interferometry.
- Traditional phase unwrapping (PU) methods fail to accurately process these deformations, resulting in unreliable monitoring.
- Precise monitoring of surface deformation in mining areas is crucial for safety and resource management.
Purpose of the Study:
- To develop a novel multi-scale phase unwrapping (PU) method for accurate deformation monitoring in mining areas.
- To enhance the robustness and accuracy of interferometric processing in the presence of large deformation gradients and noise.
- To provide a new technical solution for surface deformation inversion in large-scale mining subsidence.
Main Methods:
- Proposed a multi-scale phase unwrapping method for mining areas based on PUNet with a multi-attention mechanism (MAPUNet).
- Integrated atrous spatial pyramid pooling (ASPP) and dual attention network (DA-Net) for multi-scale interferogram feature extraction.
- Designed an efficient channel attention network (ECA-Net) optimized residual module to suppress phase noise and extract key features.
Main Results:
- MAPUNet demonstrated superior performance on simulated interferograms with high noise (SNR=1), showing a significant increase in Structural Similarity Index (SSIM).
- Compared to PUGAN, SSIM increased by 0.0176; compared to the original PUNet, SSIM increased by 0.298.
- Successfully applied MAPUNet for deformation inversion in a Shanxi mining area, detecting a maximum settlement of -9.3 cm within one month.
Conclusions:
- The proposed MAPUNet method offers higher accuracy and better robustness for phase unwrapping in mining areas with large deformation gradients.
- MAPUNet effectively addresses the challenges of monitoring large-scale subsidence in mining areas, providing results closer to real-world conditions.
- This study presents a valuable new technical reference for accurate surface deformation monitoring in mining regions.
