Deep Learning for the Detection and Phase Unwrapping of Mining-Induced Deformation in Large-Scale Interferograms

Wu, Zhipeng; Wang, Teng; Wang, Yingjie; Wang, Robert; Ge, Daqing

doi:10.1109/tgrs.2021.3121907

Cited by 31 publications

(50 citation statements)

References 73 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Particularly, Wu et al (2022) present two deep-learning networks for the aim: one is to detect mininginduced local subsidence from wrapped interferograms and the other is to unwrap the detected interferogram patches with dense fringes. However, we cannot directly adapt the strategy of Wu et al (2022) for landslide detection for two reasons as follows: one is that mining-induced subsidence is so rapid that it is apparent in single interferograms even with 6-day temporal baselines. Yet, the deformation of most landslides is slow without any visible features from single interferograms.…”

Section: Adaptation Of the Yolov3 Network Structure For Slow-moving L...mentioning

confidence: 99%

“…Yet, the deformation of most landslides is slow without any visible features from single interferograms. Second, because of their simple mechanisms, it is easy to simulate the fringe pattern produced by mining subsidence using a randomly distorted Gaussian surface as training samples (Wu et al, 2022). Yet, the deformation patterns of landslides are extremely complicated and cannot be simulated with a simple model.…”

Section: Adaptation Of the Yolov3 Network Structure For Slow-moving L...mentioning

confidence: 99%

“…With these promising applications, phase-gradient stacking shows the capability of enhancing the localized deformation signal, and a method to automatically locate these localized deformation signals is the goal next. Recently, deeplearning network not only achieved outstanding performance in computer vision fields but has also been applied in the InSAR process for mining-induced deformation (Wu et al, 2022) and generated decorrelation masks (Zhang et al, 2021), which promoted us to search for a proper network structure to detect slow-moving landslides from the phase-gradient images. Algorithms of target detection with deep learning can be divided into two categories.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Detecting slow-moving landslides using InSAR phase-gradient stacking and deep-learning network

Zhang

Wang

et al. 2022

Front. Environ. Sci.

Self Cite

View full text Add to dashboard Cite

Landslides are a major geohazard that endangers human lives and properties. Recently, efforts have been made to use Synthetic Aperture Radar Interferometry (InSAR) for landslide monitoring. However, it is still difficult to effectively and automatically identify slow-moving landslides distributed over a large area due to phase unwrapping errors, decorrelation, troposphere turbulence and computational requirements. In this study, we develop a new approach combining phase-gradient stacking and a deep-learning network based on YOLOv3 to automatically detect slow-moving landslides from large-scale interferograms. Using Sentinel-1 SAR images acquired from 2014 to 2020, we developed a burst-based, phase-gradient stacking algorithm to sum up phase gradients in short-temporal-baseline interferograms along the azimuth and range directions. The stacked phase gradients clearly reveal the characteristics of localized surface deformation that is mainly caused by slow-moving landslides and avoids the errors due to phase unwrapping in partially decorrelated areas and atmospheric effects. Then, we trained the improved Attention-YOLOv3 network with stacked phase-gradient maps of manually labeled landslides to achieve quick and automatic detection. We applied our method in an ∼180,000 km2 area of southwestern China and identified 3,366 slow-moving landslides. By comparing the results with optical imagery and previously published landslides in this region, the proposed method can achieve automatic detection over a large area precisely and efficiently. From the derived landslide density map, we determined that most landslides are distributed along the three large rivers and their branches. In addition to some counties with known high-density landslides, approximately 10 more counties with high landslide density were exposed, which should attract more attention to their risks for geohazards. This application demonstrates the potential value of our newly developed method for slow-moving landslide detection over a nation-wide area, which can be employed before applying more time-consuming time-series InSAR analysis.

show abstract

Section: Adaptation Of the Yolov3 Network Structure For Slow-moving L...mentioning

confidence: 99%

Section: Adaptation Of the Yolov3 Network Structure For Slow-moving L...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Detecting slow-moving landslides using InSAR phase-gradient stacking and deep-learning network

Zhang

Wang

et al. 2022

Front. Environ. Sci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, it is important to note that this method can be influenced by variations in threshold settings across different regions as well as the presence of data noise, lacking universality [6,[22][23][24]. Some experts focus on the automatic detection of volcano, groundsubsidence, slow-moving landslide deformation using a variety of machine learning and deep learning methods [25][26][27][28][29].…”

Section: Introductionmentioning

confidence: 99%

Active Deformation Areas of Potential Landslide Mapping with a Generalized Convolutional Neural Network

Wu,

Ge,

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

Early discovery and monitoring of the active deformation areas of potential landslides are important for geohazard risk prevention. The objective of the study is to propose a one-step strategy for automatically mapping the active deformation areas of potential landslides from a Sentinel-1 SAR dataset. First, we built a generalized convolutional neural network (CNN) based on activity and topographic characteristics. Second, we conducted a comparative analysis of the performance of various multi-channel combiners for detecting the active deformation areas of the potential landslides. Third, we verified the transferability of the pretrained CNN model for an unknown region. We found that by incorporating topographic characteristics into a generalized convolutional neural network, we were able to enhance the accuracy of identifying the active deformation areas of potential landslides, rapidly mapping these areas. The methodology is robust and efficient, and it has the capability to automatically detect the active deformation areas of potential landslides, even in unknown or unfamiliar regions. This product can facilitate automated pipelines, updating and mapping active deformation areas for final users who are not InSAR experts. This implementation can be used for providing support to risk management activities.

show abstract

“…A model-assisted phase unwrapping approach was proposed to tackle the phase aliasing problem for the cases when the subsidence gradient exceeds the maximum measurable gradient of DInSAR technique (Dai et al, 2021). Recently, a deep convolutional neural network has been proposed to detect and map localized, rapid mining subsidence from wrapped interferograms, and a phase unwrapping network is designed to unwrap the cropped interferogram patches centered on the detected subsiding locations (Wu et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

Ground subsidence associated with mining activity in the Ningdong coal base area, northwestern China revealed by InSAR time series analysis

Tang

Wang

et al. 2023

Front. Earth Sci.

View full text Add to dashboard Cite

Ningdong coal base area located in northwestern China is one of the largest coal-producing bases in China. The aim of this work is to investigate a regional-scale mining subsidence over the Ningdong coal base area, by using both conventional and advanced Differential Synthetic Aperture Radar Interferometry (DInSAR) methods. Fifteen L-band SAR images from ALOS-2 satellite and 102 C-band images from Sentinel-1A satellite spanning between November 2014 and July 2019 were used for the analysis. To increase the spatial extent of the displacement signal because of decorrelated effects, we modified the traditional Small Baseline Subset (SBAS) method to incorporate the coherence into the inverse problem, hereafter we call it coherence-based SBAS method. Instead of excluding decorrelated pixels present in the interferograms, we keep all the pixels in the time series analysis and down-weighted the decorrelated pixels with coherence. We performed the coherence-based SBAS method to both the two SAR datasets to obtain the subsidence rate maps and displacement time-series over the mining areas, and compared the results with that from the traditional stacking InSAR method. We evaluated the effectiveness of L-band and C-band DInSAR for monitoring mining subsidence by comparing differential interferograms and displacements derived from SBAS method between ALOS-2 and Sentinel-1A data. Compared to C-band, L-band SAR are less affected by phase aliasing due to large displacement gradients. The most significant subsidence was found at Maliantai mine with −264 mm/year detected by SBAS method from Sentinel-1 data. We validated the InSAR displacement accuracy by comparing both ALOS-2 and Sentinel-1 results with 18 GPS stations above five active mining regions. The average RMSE between InSAR and GPS measurements is 28.4 mm for Sentinel-1 data and 21 mm for ALOS-2 data. Our results demonstrate that the combined exploitation of L-band and C-band SAR data through both conventional and advanced DInSAR methods could be crucial to monitor ground subsidence in mining areas, which provides insights into subsidence dynamics and determine the characteristic surface response to longwall advance.

show abstract

Deep Learning for the Detection and Phase Unwrapping of Mining-Induced Deformation in Large-Scale Interferograms

Cited by 31 publications

References 73 publications

Detecting slow-moving landslides using InSAR phase-gradient stacking and deep-learning network

Detecting slow-moving landslides using InSAR phase-gradient stacking and deep-learning network

Active Deformation Areas of Potential Landslide Mapping with a Generalized Convolutional Neural Network

Ground subsidence associated with mining activity in the Ningdong coal base area, northwestern China revealed by InSAR time series analysis

Contact Info

Product

Resources

About