Building landslide inventory with LiDAR data and deep learning

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989592%3A15310%2F24%3A73628433" target="_blank" >RIV/61989592:15310/24:73628433 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1016/B978-0-12-823868-4.00014-3" target="_blank" >http://dx.doi.org/10.1016/B978-0-12-823868-4.00014-3</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/B978-0-12-823868-4.00014-3" target="_blank" >10.1016/B978-0-12-823868-4.00014-3</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Building landslide inventory with LiDAR data and deep learning

Popis výsledku v původním jazyce

The integration of data collection, cloud technologies, and artificial intelligence offers new opportunities to study Earth processes and landforms. These advances are bringing more comprehensive knowledge to the geosciences, enabling a detailed understanding of landforms and natural hazards from local to planetary scales. A shift from heuristic to data-based approaches is improving the objectivity of geoscientific analyses but is limited by data quality and availability, especially at regional scales. Landslides, influenced by geological and human factors, are an example of natural processes becoming disasters due to urbanization. Understanding past and present landslide conditions is crucial for future predictions, highlighting the importance of accurate landslide inventories. This study aims to train and evaluate deep learning models, specifically U-Net and DeepLabV3, for landslide detection and mapping in the Czech Republic using LiDAR data and landslide inventories. Results show that DeepLabV3 outperforms U-Net in accuracy, recall, and F1 score, suggesting its higher effectiveness in landslide detection. Challenges include data quality, resolution, and hillshade limitations, highlighting the need for high-quality input data for reliable AI-based geoscience applications.

Název v anglickém jazyce

Building landslide inventory with LiDAR data and deep learning

Popis výsledku anglicky

The integration of data collection, cloud technologies, and artificial intelligence offers new opportunities to study Earth processes and landforms. These advances are bringing more comprehensive knowledge to the geosciences, enabling a detailed understanding of landforms and natural hazards from local to planetary scales. A shift from heuristic to data-based approaches is improving the objectivity of geoscientific analyses but is limited by data quality and availability, especially at regional scales. Landslides, influenced by geological and human factors, are an example of natural processes becoming disasters due to urbanization. Understanding past and present landslide conditions is crucial for future predictions, highlighting the importance of accurate landslide inventories. This study aims to train and evaluate deep learning models, specifically U-Net and DeepLabV3, for landslide detection and mapping in the Czech Republic using LiDAR data and landslide inventories. Results show that DeepLabV3 outperforms U-Net in accuracy, recall, and F1 score, suggesting its higher effectiveness in landslide detection. Challenges include data quality, resolution, and hillshade limitations, highlighting the need for high-quality input data for reliable AI-based geoscience applications.

Druh

C - Kapitola v odborné knize

CEP obor

—

OECD FORD obor

10508 - Physical geography

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název knihy nebo sborníku

Earth Observation Applications to Landslide Mapping, Monitoring and Modeling

ISBN

978-0-12-823868-4

Počet stran výsledku

12

Strana od-do

297-309

Počet stran knihy

311

Název nakladatele

Elsevier Ltd.

Místo vydání

Netherlands

Kód UT WoS kapitoly

—