|
|
| Acesso ao texto completo restrito à biblioteca da Embrapa Territorial. Para informações adicionais entre em contato com cnpm.biblioteca@embrapa.br. |
Registro Completo |
Biblioteca(s): |
Embrapa Territorial. |
Data corrente: |
14/08/2020 |
Data da última atualização: |
17/08/2020 |
Tipo da produção científica: |
Artigo em Periódico Indexado |
Autoria: |
MARTINS, V. S.; KALEITA, A. L.; GELDER, B. K.; SILVEIRA, H. L. F. da; ABE, C. A. |
Afiliação: |
VITOR S. MARTINS, IOWA STATE UNIVERSITY; AMY L. KALEITA, IOWA STATE UNIVERSITY; BRIAN K. GELDER, IOWA STATE UNIVERSITY; HILTON LUIS FERRAZ DA SILVEIRA, CNPM; CAMILA A. ABE, UNIVERSITY OF WISCONSIN-MADISON. |
Título: |
Exploring multiscale object-based convolutional neural network (multi-OCNN) for remote sensing image classification at high spatial resolution. |
Ano de publicação: |
2020 |
Fonte/Imprenta: |
ISPRS Journal of Photogrammetry and Remote Sensing, v. 168, p. 56-73, oct. 2020. |
ISBN: |
0924-2716 |
DOI: |
https://doi.org/10.1016/j.isprsjprs.2020.08.004 |
Idioma: |
Inglês |
Conteúdo: |
Convolutional Neural Network (CNN) has been increasingly used for land cover mapping of remotely sensed imagery. However, large-area classification using traditional CNN is computationally expensive and produces coarse maps using a sliding window approach. To address this problem, object-based CNN (OCNN) becomes an alternative solution to improve classification performance. However, previous studies were mainly focused on urban areas or small scenes, and implementation of OCNN method is still needed for large-area classification over heterogeneous landscape. Additionally, the massive labeling of segmented objects requires a practical approach for less computation, including object analysis and multiple CNNs. This study presents a new multiscale OCNN (multi-OCNN) framework for large-scale land cover classification at 1-m resolution over 145,740 km2. Our approach consists of three main steps: (i) image segmentation, (ii) object analysis with skeleton-based algorithm, and (iii) application of multiple CNNs for final classification. Also, we developed a large benchmark dataset, called IowaNet, with 1 million labeled images and 10 classes. In our approach, multiscale CNNs were trained to capture the best contextual information during the semantic labeling of objects. Meanwhile, skeletonization algorithm provided morphological representation (?medial axis?) of objects to support the selection of convolutional locations for CNN predictions. In general, proposed multi-OCNN presented better classification accuracy (overall accuracy ~87.2%) compared to traditional patch-based CNN (81.6%) and fixed-input OCNN (82%). In addition, the results showed that this framework is 8.1 and 111.5 times faster than traditional pixel-wise CNN16 or CNN256, respectively. Multiple CNNs and object analysis have proved to be essential for accurate and fast classification. While multi-OCNN produced a high-level of spatial details in the land cover product, misclassification was observed for some classes, such as road versus buildings or shadow versus lake. Despite these minor drawbacks, our results also demonstrated the benefits of IowaNet training dataset in the model performance; overfitting process reduces as the number of samples increases. The limitations of multi-OCNN are partially explained by segmentation quality and limited number of spectral bands in the aerial data. With the advance of deep learning methods, this study supports the claim of multi-OCNN benefits for operational large-scale land cover product at 1-m resolution. MenosConvolutional Neural Network (CNN) has been increasingly used for land cover mapping of remotely sensed imagery. However, large-area classification using traditional CNN is computationally expensive and produces coarse maps using a sliding window approach. To address this problem, object-based CNN (OCNN) becomes an alternative solution to improve classification performance. However, previous studies were mainly focused on urban areas or small scenes, and implementation of OCNN method is still needed for large-area classification over heterogeneous landscape. Additionally, the massive labeling of segmented objects requires a practical approach for less computation, including object analysis and multiple CNNs. This study presents a new multiscale OCNN (multi-OCNN) framework for large-scale land cover classification at 1-m resolution over 145,740 km2. Our approach consists of three main steps: (i) image segmentation, (ii) object analysis with skeleton-based algorithm, and (iii) application of multiple CNNs for final classification. Also, we developed a large benchmark dataset, called IowaNet, with 1 million labeled images and 10 classes. In our approach, multiscale CNNs were trained to capture the best contextual information during the semantic labeling of objects. Meanwhile, skeletonization algorithm provided morphological representation (?medial axis?) of objects to support the selection of convolutional locations for CNN predictions. In general, proposed multi-OCNN presented... Mostrar Tudo |
Palavras-Chave: |
Aerial imagery; Convolutional neural network; Deep learning. |
Thesaurus Nal: |
Land cover. |
Categoria do assunto: |
-- |
Marc: |
LEADER 03374naa a2200241 a 4500 001 2124365 005 2020-08-17 008 2020 bl uuuu u00u1 u #d 022 $a0924-2716 024 7 $ahttps://doi.org/10.1016/j.isprsjprs.2020.08.004$2DOI 100 1 $aMARTINS, V. S. 245 $aExploring multiscale object-based convolutional neural network (multi-OCNN) for remote sensing image classification at high spatial resolution.$h[electronic resource] 260 $c2020 520 $aConvolutional Neural Network (CNN) has been increasingly used for land cover mapping of remotely sensed imagery. However, large-area classification using traditional CNN is computationally expensive and produces coarse maps using a sliding window approach. To address this problem, object-based CNN (OCNN) becomes an alternative solution to improve classification performance. However, previous studies were mainly focused on urban areas or small scenes, and implementation of OCNN method is still needed for large-area classification over heterogeneous landscape. Additionally, the massive labeling of segmented objects requires a practical approach for less computation, including object analysis and multiple CNNs. This study presents a new multiscale OCNN (multi-OCNN) framework for large-scale land cover classification at 1-m resolution over 145,740 km2. Our approach consists of three main steps: (i) image segmentation, (ii) object analysis with skeleton-based algorithm, and (iii) application of multiple CNNs for final classification. Also, we developed a large benchmark dataset, called IowaNet, with 1 million labeled images and 10 classes. In our approach, multiscale CNNs were trained to capture the best contextual information during the semantic labeling of objects. Meanwhile, skeletonization algorithm provided morphological representation (?medial axis?) of objects to support the selection of convolutional locations for CNN predictions. In general, proposed multi-OCNN presented better classification accuracy (overall accuracy ~87.2%) compared to traditional patch-based CNN (81.6%) and fixed-input OCNN (82%). In addition, the results showed that this framework is 8.1 and 111.5 times faster than traditional pixel-wise CNN16 or CNN256, respectively. Multiple CNNs and object analysis have proved to be essential for accurate and fast classification. While multi-OCNN produced a high-level of spatial details in the land cover product, misclassification was observed for some classes, such as road versus buildings or shadow versus lake. Despite these minor drawbacks, our results also demonstrated the benefits of IowaNet training dataset in the model performance; overfitting process reduces as the number of samples increases. The limitations of multi-OCNN are partially explained by segmentation quality and limited number of spectral bands in the aerial data. With the advance of deep learning methods, this study supports the claim of multi-OCNN benefits for operational large-scale land cover product at 1-m resolution. 650 $aLand cover 653 $aAerial imagery 653 $aConvolutional neural network 653 $aDeep learning 700 1 $aKALEITA, A. L. 700 1 $aGELDER, B. K. 700 1 $aSILVEIRA, H. L. F. da 700 1 $aABE, C. A. 773 $tISPRS Journal of Photogrammetry and Remote Sensing$gv. 168, p. 56-73, oct. 2020.
Download
Esconder MarcMostrar Marc Completo |
Registro original: |
Embrapa Territorial (CNPM) |
|
Biblioteca |
ID |
Origem |
Tipo/Formato |
Classificação |
Cutter |
Registro |
Volume |
Status |
URL |
Voltar
|
|
| Acesso ao texto completo restrito à biblioteca da Embrapa Instrumentação. Para informações adicionais entre em contato com cnpdia.biblioteca@embrapa.br. |
Registro Completo
Biblioteca(s): |
Embrapa Instrumentação. |
Data corrente: |
19/01/2006 |
Data da última atualização: |
08/09/2009 |
Autoria: |
CUNHA, T. J. F.; MATIN-NETO, L.; MADARI, B. E.; SIMÕES, M. L.; SILVA, W. T. L. da; MILORI, D. M. B. P.; BENITES, V. M.; CANELLAS, L. P.; NOVOTNY, E. H.; SANTOS, G. de A. |
Título: |
Ácidos húmicos de solos antropogênicos: estudo espectroscópico por RPE e RMN ¹³C. |
Ano de publicação: |
2005 |
Fonte/Imprenta: |
In: ENCONTRO BRASILEIRO DE SUBSTÂNCIAS HÚMICAS, 6., 2005, Rio de Janeiro. Resumos expandidos... Rio de Janeiro: EMBRAPA Solos, 2005. p. 119-122. |
Idioma: |
Português |
Conteúdo: |
Estudar as características espectroscópicas e o grau de humidificação de acidos humicos de solos com horizonte A antropico, submetidos ao uso agricola de diversas regioes da Amazonia brasileira, através de tecnicas de RPE e RMN. |
Palavras-Chave: |
EPR; Grau de humidificação; RMN. |
Thesaurus NAL: |
Amazonia. |
Categoria do assunto: |
-- |
Marc: |
LEADER 01088naa a2200277 a 4500 001 1029252 005 2009-09-08 008 2005 bl --- 0-- u #d 100 1 $aCUNHA, T. J. F. 245 $aÁcidos húmicos de solos antropogênicos$bestudo espectroscópico por RPE e RMN ¹³C. 260 $c2005 520 $aEstudar as características espectroscópicas e o grau de humidificação de acidos humicos de solos com horizonte A antropico, submetidos ao uso agricola de diversas regioes da Amazonia brasileira, através de tecnicas de RPE e RMN. 650 $aAmazonia 653 $aEPR 653 $aGrau de humidificação 653 $aRMN 700 1 $aMATIN-NETO, L. 700 1 $aMADARI, B. E. 700 1 $aSIMÕES, M. L. 700 1 $aSILVA, W. T. L. da 700 1 $aMILORI, D. M. B. P. 700 1 $aBENITES, V. M. 700 1 $aCANELLAS, L. P. 700 1 $aNOVOTNY, E. H. 700 1 $aSANTOS, G. de A. 773 $tIn: ENCONTRO BRASILEIRO DE SUBSTÂNCIAS HÚMICAS, 6., 2005, Rio de Janeiro. Resumos expandidos... Rio de Janeiro: EMBRAPA Solos, 2005. p. 119-122.
Download
Esconder MarcMostrar Marc Completo |
Registro original: |
Embrapa Instrumentação (CNPDIA) |
|
Biblioteca |
ID |
Origem |
Tipo/Formato |
Classificação |
Cutter |
Registro |
Volume |
Status |
Fechar
|
Nenhum registro encontrado para a expressão de busca informada. |
|
|