Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais
Autor(a) principal: | |
---|---|
Data de Publicação: | 2022 |
Tipo de documento: | Tese |
Idioma: | por |
Título da fonte: | Biblioteca Digital de Teses e Dissertações do UNIOESTE |
Texto Completo: | https://tede.unioeste.br/handle/tede/6136 |
Resumo: | The increase in the world population requires an expansion in food demands, consequently increasing agricultural production. Land Use and Land Cover (LULC) detailing plays an essential role in the agricultural sector, enabling efficient monitoring, planning, and management of these areas. In this segment, remote sensing techniques have proved to be a valuable tool for mapping large agricultural areas. Therefore, the general objective of this research was to explore machine learning methods to carry out the LULC mapping through satellite images of three study areas in the state of Paraná. In addition, the generalization of the models was evaluated through cross-site classification. The work was divided into three stages covered in different scientific papers. The first paper proposed a one-dimensional Temporal Convolutional Neural Network (1D-TempCNN) to classify LULC using Satellite Image Time Series (SITS). Two other classifiers, Random Forest (RF) and Support Vector Machine (SVM), were used to compare the results. The Overall Accuracy (OA) was above 98% for all models when the test was performed in the same training area. However, in the cross-site classification, 1D-TempCNN showed better OA values (between 94.34% and 98.67%) and greater generalization. Two Data Augmentation (DA) techniques, sliding window and scaling, contributed to the generalization of the models. This way, the proposed architecture proved viable for cross-site classification and can be used in different crop years (cross-year) or agricultural areas (cross-site). The second paper explored the early classification using the 1D-TempCNN architecture and two classic models, Multilayer Perceptron (MLP) and RF. The models showed similar performance, reaching OA above 95% at the end of December. However, in the cross-site classification, only the 1D-TempCNN model achieved OA above 95% in all test scenarios, reaching this value between the beginning of December and the first half of February. Thus, this model demonstrated generalization capacity and can be used for early classification in different training areas. The third paper addressed the use of semantic segmentation to build LULC maps. Two pre-trained DeepLabv3 architectures (ResNet-50 and ResNet-101) were evaluated along with two different segmentations (true color and false color) and two training image sizes (256 x 256 and 512 x 512 pixels). The reference maps used in training and testing were derived from the results of the first paper. The OA presented results between 74.91% and 77.81%, and those of the Mean Intersection over Union (MIoU) metric between 39.46% and 52.56%. In addition, the combination of false color bands was superior to true color, and the use of smaller images resulted in more detailed and accurate maps. The model with the ResNet-101 base network presented the best results in most of the analyzed metrics. However, distinguishing between soybean and corn classes was the most significant difficulty. Therefore, this model presented generalization capacity, proving to be a viable option for constructing large area LULC maps, which allows the monitoring and planning of agricultural areas. |
id |
UNIOESTE-1_f59f9a0d5799d0749d5a3a31084d6b83 |
---|---|
oai_identifier_str |
oai:tede.unioeste.br:tede/6136 |
network_acronym_str |
UNIOESTE-1 |
network_name_str |
Biblioteca Digital de Teses e Dissertações do UNIOESTE |
repository_id_str |
|
spelling |
Johann, Jerry Adrianihttp://lattes.cnpq.br/3499704308301708Camargo, Sandro da Silvahttp://lattes.cnpq.br/8826344853104147Vasata, Darlonhttp://lattes.cnpq.br/1343104664853305Catarina, Adair Santahttp://lattes.cnpq.br/7041836941307184Brun, André Luizhttp://lattes.cnpq.br/4617587198467560Opazo, Miguel Angel Uribehttp://lattes.cnpq.br/4179444121729414http://lattes.cnpq.br/6935444785347377Ló, Thiago Berticelli2022-08-09T13:06:09Z2022-05-13Ló, Thiago Berticelli. Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais. 2022. 164 f. Tese( Doutorado em Engenharia Agrícola) - Universidade Estadual do Oeste do Paraná, Cascavel, 2022.https://tede.unioeste.br/handle/tede/6136The increase in the world population requires an expansion in food demands, consequently increasing agricultural production. Land Use and Land Cover (LULC) detailing plays an essential role in the agricultural sector, enabling efficient monitoring, planning, and management of these areas. In this segment, remote sensing techniques have proved to be a valuable tool for mapping large agricultural areas. Therefore, the general objective of this research was to explore machine learning methods to carry out the LULC mapping through satellite images of three study areas in the state of Paraná. In addition, the generalization of the models was evaluated through cross-site classification. The work was divided into three stages covered in different scientific papers. The first paper proposed a one-dimensional Temporal Convolutional Neural Network (1D-TempCNN) to classify LULC using Satellite Image Time Series (SITS). Two other classifiers, Random Forest (RF) and Support Vector Machine (SVM), were used to compare the results. The Overall Accuracy (OA) was above 98% for all models when the test was performed in the same training area. However, in the cross-site classification, 1D-TempCNN showed better OA values (between 94.34% and 98.67%) and greater generalization. Two Data Augmentation (DA) techniques, sliding window and scaling, contributed to the generalization of the models. This way, the proposed architecture proved viable for cross-site classification and can be used in different crop years (cross-year) or agricultural areas (cross-site). The second paper explored the early classification using the 1D-TempCNN architecture and two classic models, Multilayer Perceptron (MLP) and RF. The models showed similar performance, reaching OA above 95% at the end of December. However, in the cross-site classification, only the 1D-TempCNN model achieved OA above 95% in all test scenarios, reaching this value between the beginning of December and the first half of February. Thus, this model demonstrated generalization capacity and can be used for early classification in different training areas. The third paper addressed the use of semantic segmentation to build LULC maps. Two pre-trained DeepLabv3 architectures (ResNet-50 and ResNet-101) were evaluated along with two different segmentations (true color and false color) and two training image sizes (256 x 256 and 512 x 512 pixels). The reference maps used in training and testing were derived from the results of the first paper. The OA presented results between 74.91% and 77.81%, and those of the Mean Intersection over Union (MIoU) metric between 39.46% and 52.56%. In addition, the combination of false color bands was superior to true color, and the use of smaller images resulted in more detailed and accurate maps. The model with the ResNet-101 base network presented the best results in most of the analyzed metrics. However, distinguishing between soybean and corn classes was the most significant difficulty. Therefore, this model presented generalization capacity, proving to be a viable option for constructing large area LULC maps, which allows the monitoring and planning of agricultural areas.O crescimento da população mundial causa uma expansão da demanda por alimentos, consequentemente, um aumento na produção agrícola. O detalhamento do uso e cobertura da terra (LULC) desempenha um papel de destaque para o setor agrícola, possibilitando o monitoramento, planejamento e gerenciamento dessas áreas de forma eficiente. Nesse segmento, o uso de técnicas de sensoriamento remoto tem se mostrado uma ferramenta valiosa para o mapeamento de extensas áreas agrícolas. Portanto, o objetivo geral da pesquisa foi explorar métodos de aprendizado de máquina visando realizar o mapeamento de LULC por meio de imagens de satélite de três áreas de estudo no estado do Paraná. Além disso, a generalização dos modelos foi avaliada por meio da classificação cruzada. Para tanto, o trabalho foi dividido em três etapas contempladas em artigos científicos. No primeiro artigo, foi proposta uma Rede Neural Convolucional Temporal Unidimensional (1D-TempCNN) para a classificação de LULC, utilizando Séries Temporais de Imagens de Satélites (SITS). Dois outros classificadores, Floresta Aleatória (RF) e Máquina de Vetores de Suporte (SVM), foram utilizados para comparação dos resultados. A Exatidão Global (EG) foi acima de 98% para todos os modelos quando o teste foi realizado na mesma área de treinamento. Entretanto, na classificação cruzada, a 1D-TempCNN apresentou melhores valores de EG (entre 94,34 e 98,67%) e maior generalização. Duas técnicas de Aumentos de Dados (DA), janela deslizante e scaling, contribuíram para generalização dos modelos. Dessa forma, a arquitetura proposta foi viável para classificação cruzada, podendo ser utilizada em diferentes anos-safra ou diferentes áreas agrícolas. No segundo artigo, foi explorada a classificação antecipada, utilizando a arquitetura 1D-TempCNN e dois modelos clássicos, Perceptron Multicamadas (MLP) e RF. Os modelos apresentaram desempenho similar, alcançando EG superior a 95% ao final de dezembro. Entretanto, na classificação cruzada, somente o modelo 1D-TempCNN alcançou EG acima de 95%, em todos os cenários testados, atingindo esse valor entre o início de dezembro e primeira quinzena de fevereiro. Dessa forma, esse modelo apresentou capacidade de generalização e pode ser utilizado para a classificação antecipada em diferentes áreas do treinamento. O terceiro artigo abordou a utilização de segmentação semântica para construção de mapas de LULC. Duas arquiteturas da Deeplabv3 (ResNet-50 e ResNet-101) pré-treinadas foram avaliadas juntamente com duas diferentes combinações de bandas (cor verdadeira e falsa cor) e dois tamanhos de imagens de treinamento. Os mapas de referência, utilizados no treinamento e teste, foram originados dos resultados do primeiro artigo. A EG apresentou resultados entre 74,91% e 77,81% e os da métrica Intersecção sobre União entre 39,46% e 52,56%. Além disso, a combinação de bandas falsa cor foi superior à de cor verdadeira, de maneira que a utilização de imagens menores resultou em mapas mais detalhados e precisos. O modelo com a rede base ResNet-101 apresentou os melhores resultados na maior parte das métricas analisadas. Contudo, a maior dificuldade encontrada foi a distinção entre as classes soja e milho. Assim, esse modelo apresentou capacidade de generalização, demostrando ser uma opção viável para a construção de mapas de LULC de área extensas, o que permitiu o monitoramento e planejamento de áreas agrícolas.Submitted by Edineia Teixeira (edineia.teixeira@unioeste.br) on 2022-08-09T13:06:09Z No. of bitstreams: 2 Thiago_Ló2022.pdf: 9844763 bytes, checksum: 332ed1ed21025cbf016668dc32d1cbae (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5)Made available in DSpace on 2022-08-09T13:06:09Z (GMT). No. of bitstreams: 2 Thiago_Ló2022.pdf: 9844763 bytes, checksum: 332ed1ed21025cbf016668dc32d1cbae (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2022-05-13application/pdfpor6588633818200016417500Universidade Estadual do Oeste do ParanáCascavelPrograma de Pós-Graduação em Engenharia AgrícolaUNIOESTEBrasilCentro de Ciências Exatas e Tecnológicashttp://creativecommons.org/licenses/by/4.0/info:eu-repo/semantics/openAccessSéries temporais de imagens de satélitesSegmentação semânticaAprendizado de máquinaGeneralizaçãoSatellite image time seriesSemantic segmentationMachine learningGeneralizationSistemas Biológicos e AgroindustriaisMapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionaisLand use and land cover mapping using remote sensing and convolutional neural networksinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesis-53476924504160521296006002214374442868382015reponame:Biblioteca Digital de Teses e Dissertações do UNIOESTEinstname:Universidade Estadual do Oeste do Paraná (UNIOESTE)instacron:UNIOESTEORIGINALThiago_Ló2022.pdfThiago_Ló2022.pdfapplication/pdf9844763http://tede.unioeste.br:8080/tede/bitstream/tede/6136/5/Thiago_L%C3%B32022.pdf332ed1ed21025cbf016668dc32d1cbaeMD55CC-LICENSElicense_urllicense_urltext/plain; charset=utf-843http://tede.unioeste.br:8080/tede/bitstream/tede/6136/2/license_url321f3992dd3875151d8801b773ab32edMD52license_textlicense_texttext/html; charset=utf-80http://tede.unioeste.br:8080/tede/bitstream/tede/6136/3/license_textd41d8cd98f00b204e9800998ecf8427eMD53license_rdflicense_rdfapplication/rdf+xml; charset=utf-80http://tede.unioeste.br:8080/tede/bitstream/tede/6136/4/license_rdfd41d8cd98f00b204e9800998ecf8427eMD54LICENSElicense.txtlicense.txttext/plain; charset=utf-82165http://tede.unioeste.br:8080/tede/bitstream/tede/6136/1/license.txtbd3efa91386c1718a7f26a329fdcb468MD51tede/61362022-08-10 10:15:33.482oai:tede.unioeste.br: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Biblioteca Digital de Teses e Dissertaçõeshttp://tede.unioeste.br/PUBhttp://tede.unioeste.br/oai/requestbiblioteca.repositorio@unioeste.bropendoar:2022-08-10T13:15:33Biblioteca Digital de Teses e Dissertações do UNIOESTE - Universidade Estadual do Oeste do Paraná (UNIOESTE)false |
dc.title.por.fl_str_mv |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
dc.title.alternative.eng.fl_str_mv |
Land use and land cover mapping using remote sensing and convolutional neural networks |
title |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
spellingShingle |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais Ló, Thiago Berticelli Séries temporais de imagens de satélites Segmentação semântica Aprendizado de máquina Generalização Satellite image time series Semantic segmentation Machine learning Generalization Sistemas Biológicos e Agroindustriais |
title_short |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
title_full |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
title_fullStr |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
title_full_unstemmed |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
title_sort |
Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais |
author |
Ló, Thiago Berticelli |
author_facet |
Ló, Thiago Berticelli |
author_role |
author |
dc.contributor.advisor1.fl_str_mv |
Johann, Jerry Adriani |
dc.contributor.advisor1Lattes.fl_str_mv |
http://lattes.cnpq.br/3499704308301708 |
dc.contributor.referee1.fl_str_mv |
Camargo, Sandro da Silva |
dc.contributor.referee1Lattes.fl_str_mv |
http://lattes.cnpq.br/8826344853104147 |
dc.contributor.referee2.fl_str_mv |
Vasata, Darlon |
dc.contributor.referee2Lattes.fl_str_mv |
http://lattes.cnpq.br/1343104664853305 |
dc.contributor.referee3.fl_str_mv |
Catarina, Adair Santa |
dc.contributor.referee3Lattes.fl_str_mv |
http://lattes.cnpq.br/7041836941307184 |
dc.contributor.referee4.fl_str_mv |
Brun, André Luiz |
dc.contributor.referee4Lattes.fl_str_mv |
http://lattes.cnpq.br/4617587198467560 |
dc.contributor.referee5.fl_str_mv |
Opazo, Miguel Angel Uribe |
dc.contributor.referee5Lattes.fl_str_mv |
http://lattes.cnpq.br/4179444121729414 |
dc.contributor.authorLattes.fl_str_mv |
http://lattes.cnpq.br/6935444785347377 |
dc.contributor.author.fl_str_mv |
Ló, Thiago Berticelli |
contributor_str_mv |
Johann, Jerry Adriani Camargo, Sandro da Silva Vasata, Darlon Catarina, Adair Santa Brun, André Luiz Opazo, Miguel Angel Uribe |
dc.subject.por.fl_str_mv |
Séries temporais de imagens de satélites Segmentação semântica Aprendizado de máquina Generalização |
topic |
Séries temporais de imagens de satélites Segmentação semântica Aprendizado de máquina Generalização Satellite image time series Semantic segmentation Machine learning Generalization Sistemas Biológicos e Agroindustriais |
dc.subject.eng.fl_str_mv |
Satellite image time series Semantic segmentation Machine learning Generalization |
dc.subject.cnpq.fl_str_mv |
Sistemas Biológicos e Agroindustriais |
description |
The increase in the world population requires an expansion in food demands, consequently increasing agricultural production. Land Use and Land Cover (LULC) detailing plays an essential role in the agricultural sector, enabling efficient monitoring, planning, and management of these areas. In this segment, remote sensing techniques have proved to be a valuable tool for mapping large agricultural areas. Therefore, the general objective of this research was to explore machine learning methods to carry out the LULC mapping through satellite images of three study areas in the state of Paraná. In addition, the generalization of the models was evaluated through cross-site classification. The work was divided into three stages covered in different scientific papers. The first paper proposed a one-dimensional Temporal Convolutional Neural Network (1D-TempCNN) to classify LULC using Satellite Image Time Series (SITS). Two other classifiers, Random Forest (RF) and Support Vector Machine (SVM), were used to compare the results. The Overall Accuracy (OA) was above 98% for all models when the test was performed in the same training area. However, in the cross-site classification, 1D-TempCNN showed better OA values (between 94.34% and 98.67%) and greater generalization. Two Data Augmentation (DA) techniques, sliding window and scaling, contributed to the generalization of the models. This way, the proposed architecture proved viable for cross-site classification and can be used in different crop years (cross-year) or agricultural areas (cross-site). The second paper explored the early classification using the 1D-TempCNN architecture and two classic models, Multilayer Perceptron (MLP) and RF. The models showed similar performance, reaching OA above 95% at the end of December. However, in the cross-site classification, only the 1D-TempCNN model achieved OA above 95% in all test scenarios, reaching this value between the beginning of December and the first half of February. Thus, this model demonstrated generalization capacity and can be used for early classification in different training areas. The third paper addressed the use of semantic segmentation to build LULC maps. Two pre-trained DeepLabv3 architectures (ResNet-50 and ResNet-101) were evaluated along with two different segmentations (true color and false color) and two training image sizes (256 x 256 and 512 x 512 pixels). The reference maps used in training and testing were derived from the results of the first paper. The OA presented results between 74.91% and 77.81%, and those of the Mean Intersection over Union (MIoU) metric between 39.46% and 52.56%. In addition, the combination of false color bands was superior to true color, and the use of smaller images resulted in more detailed and accurate maps. The model with the ResNet-101 base network presented the best results in most of the analyzed metrics. However, distinguishing between soybean and corn classes was the most significant difficulty. Therefore, this model presented generalization capacity, proving to be a viable option for constructing large area LULC maps, which allows the monitoring and planning of agricultural areas. |
publishDate |
2022 |
dc.date.accessioned.fl_str_mv |
2022-08-09T13:06:09Z |
dc.date.issued.fl_str_mv |
2022-05-13 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/doctoralThesis |
format |
doctoralThesis |
status_str |
publishedVersion |
dc.identifier.citation.fl_str_mv |
Ló, Thiago Berticelli. Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais. 2022. 164 f. Tese( Doutorado em Engenharia Agrícola) - Universidade Estadual do Oeste do Paraná, Cascavel, 2022. |
dc.identifier.uri.fl_str_mv |
https://tede.unioeste.br/handle/tede/6136 |
identifier_str_mv |
Ló, Thiago Berticelli. Mapeamento de uso e cobertura da terra utilizando sensoriamento remoto e redes neurais convolucionais. 2022. 164 f. Tese( Doutorado em Engenharia Agrícola) - Universidade Estadual do Oeste do Paraná, Cascavel, 2022. |
url |
https://tede.unioeste.br/handle/tede/6136 |
dc.language.iso.fl_str_mv |
por |
language |
por |
dc.relation.program.fl_str_mv |
-5347692450416052129 |
dc.relation.confidence.fl_str_mv |
600 600 |
dc.relation.department.fl_str_mv |
2214374442868382015 |
dc.rights.driver.fl_str_mv |
http://creativecommons.org/licenses/by/4.0/ info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
http://creativecommons.org/licenses/by/4.0/ |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.publisher.none.fl_str_mv |
Universidade Estadual do Oeste do Paraná Cascavel |
dc.publisher.program.fl_str_mv |
Programa de Pós-Graduação em Engenharia Agrícola |
dc.publisher.initials.fl_str_mv |
UNIOESTE |
dc.publisher.country.fl_str_mv |
Brasil |
dc.publisher.department.fl_str_mv |
Centro de Ciências Exatas e Tecnológicas |
publisher.none.fl_str_mv |
Universidade Estadual do Oeste do Paraná Cascavel |
dc.source.none.fl_str_mv |
reponame:Biblioteca Digital de Teses e Dissertações do UNIOESTE instname:Universidade Estadual do Oeste do Paraná (UNIOESTE) instacron:UNIOESTE |
instname_str |
Universidade Estadual do Oeste do Paraná (UNIOESTE) |
instacron_str |
UNIOESTE |
institution |
UNIOESTE |
reponame_str |
Biblioteca Digital de Teses e Dissertações do UNIOESTE |
collection |
Biblioteca Digital de Teses e Dissertações do UNIOESTE |
bitstream.url.fl_str_mv |
http://tede.unioeste.br:8080/tede/bitstream/tede/6136/5/Thiago_L%C3%B32022.pdf http://tede.unioeste.br:8080/tede/bitstream/tede/6136/2/license_url http://tede.unioeste.br:8080/tede/bitstream/tede/6136/3/license_text http://tede.unioeste.br:8080/tede/bitstream/tede/6136/4/license_rdf http://tede.unioeste.br:8080/tede/bitstream/tede/6136/1/license.txt |
bitstream.checksum.fl_str_mv |
332ed1ed21025cbf016668dc32d1cbae 321f3992dd3875151d8801b773ab32ed d41d8cd98f00b204e9800998ecf8427e d41d8cd98f00b204e9800998ecf8427e bd3efa91386c1718a7f26a329fdcb468 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 MD5 MD5 MD5 |
repository.name.fl_str_mv |
Biblioteca Digital de Teses e Dissertações do UNIOESTE - Universidade Estadual do Oeste do Paraná (UNIOESTE) |
repository.mail.fl_str_mv |
biblioteca.repositorio@unioeste.br |
_version_ |
1801124589760151552 |