Application of the YOLOv11-seg algorithm for AI-based landslide detection and recognition

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 11, 2025

In recent years, landslides have occurred frequently around the world, resulting in significant casualties and property damage. A notable example 2014, when a landslide Argo region of Afghanistan claimed over 2000 lives, becoming one most devastating events history. The increasing frequency severity present challenges to geological disaster monitoring, making development efficient accurate detection methods critical for mitigation prevention. This study proposes an intelligent recognition method landslides, which is based on latest deep learning model, YOLOv11-seg, designed address posed by complex terrains diverse characteristics landslides. Using Bijie-Landslide dataset, optimizes feature extraction segmentation modules enhancing both accuracy boundary pixel-level areas. Compared with traditional methods, YOLOv11-seg performs better detecting boundaries handling occlusion, demonstrating superior quality. During preprocessing phase, various data augmentation techniques, including mirroring, rotation, color adjustment, were employed, significantly improving model's generalization performance robustness across varying terrains, seasons, lighting conditions. experimental results indicate that model excels several key metrics, such as precision, recall, F1 score, mAP. Specifically, score reaches 0.8781 0.8114 segmentation, whereas mAP bounding box (B) mask (M) tasks outperforms methods. These highlight high reliability adaptability detection. research provides new technological support monitoring risk assessment, highlighting its potential monitoring.

Language: Английский

---

Satellite‐Aided Disaster Response

AGU Advances, Journal Year: 2025, Volume and Issue: 6(1)

Published: Feb. 1, 2025

Abstract The increasing frequency and severity of natural disasters, driven by climate change anthropogenic activities, pose unprecedented challenges to emergency response agencies worldwide. Satellite remote sensing has become a critical tool for providing timely accurate data aid in disaster preparedness, response, recovery. This Commentary explores the role satellite managing climate‐driven highlighting use technologies such as Synthetic Aperture Radar (SAR) creating damage proxy maps. These maps are instrumental assessing impacts guiding efforts, demonstrated 2023 Wildfires Hawaii. Despite promise these tools, remain, including need rapid processing, automation pipelines, robust international collaborations. future missions composing Earth System Observatory, upcoming NASA‐ISRO SAR mission, represents significant advancement with its global coverage frequent, detailed measurements. study emphasizes importance continued investment advanced cooperation enhance capabilities, ultimately building more resilient community.

Language: Английский

---

A Novel Multi‐Layer Attention Boosted YOLOv10 Network for Landslide Mapping Using Remote Sensing Data

Transactions in GIS, Journal Year: 2025, Volume and Issue: 29(2)

Published: March 9, 2025

ABSTRACT Detecting landslides is a critical challenge within the remote sensing fraternity, especially given need for timely and accurate hazard assessment. Traditional methods identifying from data are often manual or partially automated; however, with progress of computer vision technology, automated based on deep learning algorithms have gained significant attention. Furthermore, attention mechanisms, inspired by human visual structure, grown remarkably in various applications, including studies. In this study, we leverage capabilities YOLO models, YOLOv10 its variants, to automate detection landslides. We applied four prevailing mechanisms: CBAM, ECA, GAM, SA. Models trained using Bijie landslide database. Moreover, best results unveiled evaluation criteria, that is, precision, recall, f‐score, mAP. The YOLOv10m+CBAM showed performance map@50‐95 78.5%. Our demonstrate robust system capable rapidly localizing events speed accuracy improvements. This advancement augments process supports more effective disaster response management.

Language: Английский

---

Dynamic failure mechanisms and hazard evaluation of rock collapse induced by extreme rainfall in Changbai County highways

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 21, 2025

Rock collapses induced by extreme rainfall frequently occur along highways in Changbai County, posing serious threats to traffic safety and regional sustainable development. This study introduces a slope-unit zoning approach into the hazard assessment of collapses, integrating UDEC (Universal Distinct Element Code) numerical simulation GIS (Geographic Information System) technology reveal failure mechanism affected areas slopes under conditions. By employing AHP-CV (Analytic Hierarchy Process-Coefficient Variation) combined weighting method, weights nine critical indicators, including elevation, slope, slope direction, NDVI (Normalized Difference Vegetation Index), were quantified. Pearson Type III frequency analysis was used estimate recurrence periods, collapse distribution different probabilities evaluated. The results indicate that extremely high susceptibility are primarily distributed steep with fault development sparse vegetation, accounting for 19.74% total area. Under 100-year return condition, proportion high-hazard increases 38.68%. Increased pore water pressure reduced shear strength joint planes identified as primary causes tensile-collapse composite slopes. model achieved an AUC value 0.908, demonstrating reliability. overcomes limitations traditional grid-unit methods provides scientific insights technical support analysis, assessment, prevention geological disasters

Language: Английский

---

The Predictive Skill of a Remote Sensing-Based Machine Learning Model for Ice Wedge and Visible Ground Ice Identification in Western Arctic Canada

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(7), P. 1245 - 1245

Published: April 1, 2025

Fine-scale maps of ground ice and related surface features are critical for permafrost-related modelling management. However, such lacking across almost the entire Arctic. Machine learning provides potential to automate regional fine-scale mapping using remote sensing topographic data. Here, we evaluate predictive skill XGBoost models identifying (1) wedge (2) top-5m visible in Tuktoyaktuk Coastlands. We find high occurrence (ROC AUC = 0.95, macro F1 0.80), with most important predictors being slope, distance coast, probability depression. The model accurately predicted local trends occurrence, an increase polygon (IWP) towards coast poorly drained depressions. also captured IWP well-drained uplands study area, including locations troughs not contained training Spatial transferability analyses highlight variability probability, reflecting contrasting climatic conditions. Conversely, low 0.67, 0.53) is attributed limitations data weak associations remotely sensed predictors. varying accuracy highlights importance high-quality reference site-specific conditions improving studies data-driven from observations.

Language: Английский

---

Prediction Method of Slope Sliding Long‐Term Deformation Considering Rainfall

Advances in Civil Engineering, Journal Year: 2025, Volume and Issue: 2025(1)

Published: Jan. 1, 2025

Despite extensive research on slope seepage mechanisms, a reliable long‐term prediction method for deformation considering rainfall remains undeveloped, largely due to the complexity of rainfall‐induced instability. This study leverages project in engineering explore under heavy using intelligent monitoring techniques and genetic algorithm (GA) optimization neural network prediction. By analyzing patterns varied intensities, results reveal that limited has minimal impact stability, whereas excessive disrupts internal patterns, increasing pore water pressure reducing soil shear strength, it thereby enhances risk instability potential landslides, significantly impacting stability. The GA‐optimized accurately captures abrupt stages, avoids local optima, provides viable framework early warning

Language: Английский

---

Landslide detection based on pixel-level contrastive learning for semi-supervised semantic segmentation in wide areas

Landslides, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 11, 2024

Language: Английский

---

A Novel Framework for Spatiotemporal Susceptibility Prediction of Rainfall-Induced Landslides: A Case Study in Western Pennsylvania

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(18), P. 3526 - 3526

Published: Sept. 23, 2024

Landslide susceptibility measures the probability of landslides occurring under certain geo-environmental conditions and is essential in landslide hazard assessment. mapping (LSM) using data-driven methods applies statistical models geospatial data to show relative propensity slope failure a given area. However, due rarity multi-temporal inventory, conventional LSMs are primarily generated by spatial causative factors, while temporal factors remain limited. In this study, spatiotemporal LSM carried out machine learning (ML) techniques assess rainfall-induced susceptibility. To achieve this, two inventories collected for southwestern Pennsylvania: inventory with 4543 223 historical samples, respectively. The lacks information describe distribution; there insufficient samples represent distribution. A novel paradigm augmentation through non-landslide sampling based on domain knowledge applied leverage both ML modeling. results that model proposed predicts well space time across study area, value 0.86 area receiver operating characteristic curve (AUC), which makes it an effective tool mitigation forecasting.

Language: Английский

---

A novel framework for debris flow susceptibility assessment considering the uncertainty of sample selection

Geomatics Natural Hazards and Risk, Journal Year: 2024, Volume and Issue: 15(1)

Published: Nov. 10, 2024

The uncertainty arising from random sampling of non-debris flow samples significantly impacts the accuracy debris susceptibility assessments (DFSA). This study introduces a novel elimination method, Kernel Density Estimation (KDE), and compares it with Mean Maximum Probability Analysis (MPA) methods. Furthermore, we investigate responses four commonly used machine learning models to uncertainty, comparing two structurally similar (Random Forest (RF) Extremely Randomized Trees (ERT)) different (Support Vector Machine (SVM) Multilayer Perceptron (MLP)). results indicate that application these methods can enhance AUC values zoning accuracy, KDE method outperforming others. Specifically, based on for RF, ERT, SVM, MLP are 0.995, 0.999, 0.853, respectively. corresponding is 1.00, 0.78, further reveals vary by model architecture: SVM typically exhibit bimodal normal distributions, while shows unimodal distribution. Additionally, more sensitive variations in negative samples, whereas RF ERT less affected due ensemble structure.

Language: Английский

---