Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ Repositório do ISCTE...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
addClaim

Segmentação de imagens de zonas urbanas em espaços florestais

Authors: Dinis, Diogo Alexandre Ferreira;

Segmentação de imagens de zonas urbanas em espaços florestais

Abstract

Os incêndios florestais são das catástrofes naturais mais graves, não apenas pela frequência com que estes acontecem e dimensão que alcançam, como também pelos efeitos devastadores económicos e ambientais. Nos últimos anos, Portugal não tem tido a capacidade de travar eficazmente estes fenómenos naturais. Torna-se por isso necessário concentrar esforços na prevenção e na deteção, de forma que se consiga uma ação mais rápida sobre os focos de incêndios florestais assim que estes emergem. O INOV (Instituto de Engenharia de Sistemas e Computadores Inovação) desenvolveu um sistema de televigilância – CICLOPE – para monitorização de incêndios florestais através de torres de videovigilância que recolhem imagens. Entre as funcionalidades implementadas, existe um sistema responsável pela deteção automática de incêndios. Contudo, para esse sistema funcionar corretamente, as imagens recolhidas necessitam de estar devidamente segmentadas entre espaços florestais e zonas urbanas. Esta segmentação é realizada manualmente pelos profissionais do INOV, sendo um processo longo e suscetível a erros. Com o intuito de automatizar o processo de segmentação, nesta dissertação foi elaborada uma base de dados recorrendo às imagens fornecidas pelo INOV. Após a análise dos dados obtidos foram estudadas várias abordagens de segmentação, sendo selecionadas três que melhor se adequavam ao problema. As abordagens selecionadas foram, thresholding, K-Means e a arquitetura DeepLabv3, sendo as últimas duas referentes à vertente de aprendizagem automática. Após uma comparação dos resultados dos três modelos estudados, a arquitetura DeepLabv3 exibiu o melhor resultado, apresentando os valores médios de Intersection over Union e Dice Coefficient 0,51 e 0,78, respetivamente.

Forest fires are among the most serious natural disasters, not only because of their frequency and scale, but also because of their devastating economic and environmental effects. In the last years, Portugal has not had the capacity to effectively stop this natural phenomenon. It is therefore necessary to concentrate efforts on prevention and detection, so that faster action can be taken on forest fire outbreaks as soon as they emerge. INOV (Instituto de Engenharia de Sistemas e Computadores Inovação) has developed a surveillance system - CICLOPE - for monitoring forest fires through video surveillance towers that collect images. Among the implemented functionalities, there is a system responsible for automatic fire detection. However, for this system to work properly, the collected images need to be properly segmented between forest spaces and urban areas to minimize the number of false positives. This segmentation is performed manually by INOV professionals and is a long process and susceptible to errors. With the aim of automating the segmentation process, in this dissertation a database was developed using the images provided by INOV. After the analysis of the obtained data, several approaches of segmentation were studied, being selected three that best suited the problem. The approaches selected were thresholding, K-Means and DeepLabv3 architecture, the last two referring to machine learning. After a comparison of the results of the three models studied, the DeepLabv3 architecture exhibited the best result, presenting the average values of Intersection over Union and Dice Coefficient 0,51 and 0,78, respectively.

Country
Portugal
Keywords

Image segmentation, Domínio/Área Científica::Ciências Sociais::Economia e Gestão, Segmentação de imagem, Incêndios florestais, Rede neuronal convolucional - -- Convolutional neural network (CNN or ConvNet), Visão computacional -- Computer vision, Wildfires

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
    OpenAIRE UsageCounts
    Usage byUsageCounts
    visibility views 49
    download downloads 47
  • 49
    views
    47
    downloads
    Powered byOpenAIRE UsageCounts
Powered by OpenAIRE graph
Found an issue? Give us feedback
visibility
download
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
views
OpenAIRE UsageCountsViews provided by UsageCounts
downloads
OpenAIRE UsageCountsDownloads provided by UsageCounts
0
Average
Average
Average
49
47
Green
Related to Research communities