Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images
The availability of remote-sensing multisource data from optical-based satellite sensors has created new opportunities and challenges for forest monitoring in the Amazon Biome. In particular, change-detection analysis has emerged in recent decades to monitor forest-change dynamics, supporting some B...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2021-12-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/13/24/5084 |
_version_ | 1797501007458992128 |
---|---|
author | Daliana Lobo Torres Javier Noa Turnes Pedro Juan Soto Vega Raul Queiroz Feitosa Daniel E. Silva Jose Marcato Junior Claudio Almeida |
author_facet | Daliana Lobo Torres Javier Noa Turnes Pedro Juan Soto Vega Raul Queiroz Feitosa Daniel E. Silva Jose Marcato Junior Claudio Almeida |
author_sort | Daliana Lobo Torres |
collection | DOAJ |
description | The availability of remote-sensing multisource data from optical-based satellite sensors has created new opportunities and challenges for forest monitoring in the Amazon Biome. In particular, change-detection analysis has emerged in recent decades to monitor forest-change dynamics, supporting some Brazilian governmental initiatives such as PRODES and DETER projects for biodiversity preservation in threatened areas. In recent years fully convolutional network architectures have witnessed numerous proposals adapted for the change-detection task. This paper comprehensively explores state-of-the-art fully convolutional networks such as U-Net, ResU-Net, SegNet, FC-DenseNet, and two DeepLabv3+ variants on monitoring deforestation in the Brazilian Amazon. The networks’ performance is evaluated experimentally in terms of Precision, Recall, <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>F</mi><mn>1</mn></mrow></semantics></math></inline-formula>-score, and computational load using satellite images with different spatial and spectral resolution: Landsat-8 and Sentinel-2. We also include the results of an unprecedented auditing process performed by senior specialists to visually evaluate each deforestation polygon derived from the network with the highest accuracy results for both satellites. This assessment allowed estimation of the accuracy of these networks simulating a process “in nature” and faithful to the PRODES methodology. We conclude that the high resolution of Sentinel-2 images improves the segmentation of deforestation polygons both quantitatively (in terms of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>F</mi><mn>1</mn></mrow></semantics></math></inline-formula>-score) and qualitatively. Moreover, the study also points to the potential of the operational use of Deep Learning (DL) mapping as products to be consumed in PRODES. |
first_indexed | 2024-03-10T03:12:04Z |
format | Article |
id | doaj.art-e7a5eb357c744f879f45a66819b23a33 |
institution | Directory Open Access Journal |
issn | 2072-4292 |
language | English |
last_indexed | 2024-03-10T03:12:04Z |
publishDate | 2021-12-01 |
publisher | MDPI AG |
record_format | Article |
series | Remote Sensing |
spelling | doaj.art-e7a5eb357c744f879f45a66819b23a332023-11-23T10:24:33ZengMDPI AGRemote Sensing2072-42922021-12-011324508410.3390/rs13245084Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 ImagesDaliana Lobo Torres0Javier Noa Turnes1Pedro Juan Soto Vega2Raul Queiroz Feitosa3Daniel E. Silva4Jose Marcato Junior5Claudio Almeida6Department of Electrical Engineering, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro 22451-900, BrazilDepartment of Electrical Engineering, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro 22451-900, BrazilDepartment of Electrical Engineering, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro 22451-900, BrazilDepartment of Electrical Engineering, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro 22451-900, BrazilInstituto Nacional de Pesquisas Espaciais—INPE, São José dos Campos 12227-010, BrazilFaculty of Engineering, Architecture and Urbanism and Geography, Federal University of Mato Grosso do Sul, Campo Grande 79070-900, BrazilInstituto Nacional de Pesquisas Espaciais—INPE, São José dos Campos 12227-010, BrazilThe availability of remote-sensing multisource data from optical-based satellite sensors has created new opportunities and challenges for forest monitoring in the Amazon Biome. In particular, change-detection analysis has emerged in recent decades to monitor forest-change dynamics, supporting some Brazilian governmental initiatives such as PRODES and DETER projects for biodiversity preservation in threatened areas. In recent years fully convolutional network architectures have witnessed numerous proposals adapted for the change-detection task. This paper comprehensively explores state-of-the-art fully convolutional networks such as U-Net, ResU-Net, SegNet, FC-DenseNet, and two DeepLabv3+ variants on monitoring deforestation in the Brazilian Amazon. The networks’ performance is evaluated experimentally in terms of Precision, Recall, <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>F</mi><mn>1</mn></mrow></semantics></math></inline-formula>-score, and computational load using satellite images with different spatial and spectral resolution: Landsat-8 and Sentinel-2. We also include the results of an unprecedented auditing process performed by senior specialists to visually evaluate each deforestation polygon derived from the network with the highest accuracy results for both satellites. This assessment allowed estimation of the accuracy of these networks simulating a process “in nature” and faithful to the PRODES methodology. We conclude that the high resolution of Sentinel-2 images improves the segmentation of deforestation polygons both quantitatively (in terms of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>F</mi><mn>1</mn></mrow></semantics></math></inline-formula>-score) and qualitatively. Moreover, the study also points to the potential of the operational use of Deep Learning (DL) mapping as products to be consumed in PRODES.https://www.mdpi.com/2072-4292/13/24/5084Amazon biomechange detectiondeep learningfully convolutional neural networksremote sensingsemantic segmentation |
spellingShingle | Daliana Lobo Torres Javier Noa Turnes Pedro Juan Soto Vega Raul Queiroz Feitosa Daniel E. Silva Jose Marcato Junior Claudio Almeida Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images Remote Sensing Amazon biome change detection deep learning fully convolutional neural networks remote sensing semantic segmentation |
title | Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images |
title_full | Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images |
title_fullStr | Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images |
title_full_unstemmed | Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images |
title_short | Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images |
title_sort | deforestation detection with fully convolutional networks in the amazon forest from landsat 8 and sentinel 2 images |
topic | Amazon biome change detection deep learning fully convolutional neural networks remote sensing semantic segmentation |
url | https://www.mdpi.com/2072-4292/13/24/5084 |
work_keys_str_mv | AT dalianalobotorres deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images AT javiernoaturnes deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images AT pedrojuansotovega deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images AT raulqueirozfeitosa deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images AT danielesilva deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images AT josemarcatojunior deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images AT claudioalmeida deforestationdetectionwithfullyconvolutionalnetworksintheamazonforestfromlandsat8andsentinel2images |