Using the U-net convolutional network to map forest types and disturbance in the Atlantic rainforest with very high resolution images

Autor(es)

Wagner, Fabien H. ^[1] ; Sanchez, Alber ^[1] ; Tarabalka, Yuliya ^{[2, 3]} ; Lotte, Rodolfo G. ^[1] ; Ferreira, Matheus P. ^{[1, 4]} ; Aidar, Marcos P. M. ^[5] ; Gloor, Emanuel ^[6] ; Phillips, Oliver L. ^[6] ; Aragao, Luiz E. O. C. ^{[1, 7]}

Afiliação do(s) autor(es)

^[1] Natl Inst Space Res INPE, Remote Sensing Div, BR-12227010 Sao Jose Dos Campos, SP - Brazil

^[2] INRIA Sophia Antipolis, F-06902 Sophia Antipolis - France

^[3] Luxcarta Technol, Parc Act Argile, Lot 119b, F-06370 Mouans Sartoux - France

^[4] Mil Inst Engn IME, Cartog Engn Sect, Praca Gen Tiburcio 80, BR-22290270 Rio De Janeiro, RJ - Brazil

^[5] Inst Bot, Dept Plant Physiol & Biochem, PB 4005, BR-01061970 Sao Paulo - Brazil

^[6] Univ Leeds, Sch Geog, Ecol & Global Change, Leeds LS2 9JT, W Yorkshire - England

^[7] Univ Exeter, Coll Life & Environm Sci, Exeter EX4 4RJ, Devon - England

• Tipo de documento: Artigo Científico
• Fonte: REMOTE SENSING IN ECOLOGY AND CONSERVATION; v. 5, n. 4, p. 360-375, DEC 2019.
• Citações Web of Science: 17

Resumo

Mapping forest types and tree species at regional scales to provide information for ecologists and forest managers is a new challenge for the remote sensing community. Here, we assess the potential of a U-net convolutional network, a recent deep learning algorithm, to identify and segment (1) natural forests and eucalyptus plantations, and (2) an indicator of forest disturbance, the tree species Cecropia hololeuca, in very high resolution images (0.3 m) from the WorldView-3 satellite in the Brazilian Atlantic rainforest region. The networks for forest types and Cecropia trees were trained with 7611 and 1568 red-green-blue (RGB) images, respectively, and their dense labeled masks. Eighty per cent of the images were used for training and 20% for validation. The U-net network segmented forest types with an overall accuracy >95% and an intersection over union (IoU) of 0.96. For C. hololeuca, the overall accuracy was 97% and the IoU was 0.86. The predictions were produced over a 1600 km(2) region using WorldView-3 RGB bands pan-sharpened at 0.3 m. Natural and eucalyptus forests compose 79 and 21% of the region's total forest cover (82 250 ha). Cecropia crowns covered 1% of the natural forest canopy. An index to describe the level of disturbance of the natural forest fragments based on the spatial distribution of Cecropia trees was developed. Our work demonstrates how a deep learning algorithm can support applications such as vegetation, tree species distributions and disturbance mapping on a regional scale. (AU)

• Processo FAPESP: 16/03397-7 - Ferramentas para gerenciar imagens de satélite em array databases
• Beneficiário: Alber Hamersson Sánchez Ipia
• Modalidade de apoio: Bolsas no Brasil - Programa Capacitação - Treinamento Técnico
• Processo FAPESP: 16/24977-1 - Mapeamento e extração de atributos químicos de espécies arbóreas da Mata Atlântica por meio de imagens hiperespectrais e modelagem de transferência radiativa
• Beneficiário: Matheus Pinheiro Ferreira
• Modalidade de apoio: Bolsas no Brasil - Pós-Doutorado
• Processo FAPESP: 15/50484-0 - Diversidade funcional dos biomas Amazônia, Mata Atlântica e Cerrado nos ambientes intactos e em regeneração por meio de imagens hiperspectrais
• Beneficiário: Fabien Hubert Wagner
• Modalidade de apoio: Auxílio à Pesquisa - Jovens Pesquisadores
• Processo FAPESP: 13/50533-5 - Entendendo as respostas do metabolismo fotossintético a variações climáticas sazonais em florestas tropicais
• Beneficiário: Luiz Eduardo Oliveira e Cruz de Aragão
• Modalidade de apoio: Auxílio à Pesquisa - Regular
• Processo FAPESP: 12/51872-5 - ECOFOR: Biodiversidade e funcionamento de ecossistemas em áreas alteradas pelo homem nas Florestas Amazônica e Atlântica
• Beneficiário: Carlos Alfredo Joly
• Modalidade de apoio: Auxílio à Pesquisa - Programa BIOTA - Temático