Related Experiment Video
Updated: Jun 20, 2026
07:58
Data Processing Methods for 3D Seismic Imaging of Subsurface Volcanoes: Applications to the Tarim Flood Basalt
Published on: August 7, 2017
10.0K
A Multi-Scale Global Fusion-Based Method for Surface Fissure Extraction from UAV Imagery.
Mingxi Zhou1, Min Ji1,2,3,4, Fengxiang Jin1,2
1College of Geodesy and Geomatics, Shandong University of Science and Technology, Qingdao 266590, China.
Sensors (Basel, Switzerland)
|March 14, 2026
Summary
A new MGF-UNet model accurately extracts ground fissures, crucial for geohazard monitoring and environmental protection in deforming areas. This method enhances safety and aids ecological restoration efforts.
Area of Science:
- Geosciences and Remote Sensing
- Artificial Intelligence in Earth Observation
Background:
- Ground fissures in deforming areas pose significant risks to operational safety and the natural environment.
- Elongated morphologies and large-scale variations of fissures present challenges for accurate feature extraction.
Purpose of the Study:
- To develop an advanced semantic segmentation network, MGF-UNet, for accurate automatic fissure extraction.
- To improve the detection and characterization of ground fissures in complex terrains.
Main Methods:
- Integration of multi-scale feature sensing (MFS) and grouped efficient multi-scale attention (EMA) in shallow layers.
- Utilizing a Token-Selective Context Transformer (TSCT) for selective global modeling of semantic features in deeper layers.
- Employing feature-wise linear modulation (FiLM) for cross-level interaction and a Fourier transform-based adaptive feature fusion (AFF) module in the decoder.
Main Results:
- MGF-UNet achieved 78.2% accuracy, 81.4% Dice score, and 68.6% IoU on benchmark datasets (MFD and road-based).
- The proposed network outperformed existing mainstream networks in fissure extraction tasks.
- Demonstrated effectiveness in suppressing background noise and enhancing boundary contrast for elongated fissures.
Conclusions:
- MGF-UNet offers an effective solution for automatic fissure extraction in deformation-prone environments.
- The method has significant potential for enhancing geohazard monitoring and supporting ecological restoration initiatives.
