Automatic delimitation of volcanic ash in satellite images using Deep Learning

Artificial Intelligence has had a big impact in recent years, this field of Informatics is increasingly used to solve geological problems. One of the main applications is the detection and segmentation of volcanic ash in satellite images. For this purpose, we propose a Deep Learning model based on a...

ver descrição completa

Detalhes bibliográficos
Autores: Aldás-Núñez, Roberth Joel, Tuz-Chamorro, Katherin Vanessa, Vega-Ocaña, Jair Alejandro, Velasco-Haro, Marco Sebastián, Mejía-Escobar, Christian Iván
Tipo de documento: artigo
Estado:Versão publicada
Data de publicação:2022
País:Ecuador
Recursos:Universidad Central del Ecuador
Repositório:Revista FIGEMPA: Investigación y Desarrollo
Idioma:espanhol
OAI Identifier:oai:revistadigital.uce.edu.ec:article/3121
Acesso em linha:https://revistadigital.uce.edu.ec/index.php/RevFIG/article/view/3121
Access Level:Acceso aberto
Palavra-chave:deep learning
red neuronal convolucional
imágenes satelitales
volcán sangay
segmentación de ceniza
convolutional neuronal network
satellite images
sangay volcano
ash segmentation
Descrição
Resumo:Artificial Intelligence has had a big impact in recent years, this field of Informatics is increasingly used to solve geological problems. One of the main applications is the detection and segmentation of volcanic ash in satellite images. For this purpose, we propose a Deep Learning model based on a Convolutional Neural Network (CNN), trained with a satellite image dataset where the "ash" filter is applied, which provides a reddish-pink coloration to the ash, facilitating the segmentation process. The results show an accuracy of 99%, which is suitable for the segmentation of the ash emitted by Sangay Volcano, which has presented periods of volcanic activity in recent years. Our model generated segmented images that are consistent with the studies published by the IG-EPN.