Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems
The Sentinel-2 satellite mission offers high resolution multispectral time-series image data, enabling the production of detailed land cover maps globally. When mapping large territories, the trade-off between processing time and result quality is a central design decision. Currently, this machine l...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2019-08-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/11/17/1986 |
id |
doaj-1005128531594afb9fef5dbfd38e73f7 |
---|---|
record_format |
Article |
spelling |
doaj-1005128531594afb9fef5dbfd38e73f72020-11-25T02:52:35ZengMDPI AGRemote Sensing2072-42922019-08-011117198610.3390/rs11171986rs11171986Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational SystemsAndrei Stoian0Vincent Poulain1Jordi Inglada2Victor Poughon3Dawa Derksen4Thales/SIX/ThereSiS, 91477 Palaiseau, FranceThales Services, 31555 Toulouse, FranceCentre d’Etudes Spatiales de la BIOsphere (CESBIO), Université de Toulouse, CNES/CNRS/IRD/UPS/INRA, 31555 Toulouse, FranceCentre National d’Etudes Spatiales (CNES), 31555 Toulouse, FranceCentre d’Etudes Spatiales de la BIOsphere (CESBIO), Université de Toulouse, CNES/CNRS/IRD/UPS/INRA, 31555 Toulouse, FranceThe Sentinel-2 satellite mission offers high resolution multispectral time-series image data, enabling the production of detailed land cover maps globally. When mapping large territories, the trade-off between processing time and result quality is a central design decision. Currently, this machine learning task is usually performed using pixel-wise classification methods. However, the radical shift of the computer vision field away from hand-engineered image features and towards more automation by representation learning comes with many promises, including higher quality results and less engineering effort. In particular, convolutional neural networks learn features which take into account the context of the pixels and, therefore, a better representation of the data can be obtained. In this paper, we assess fully convolutional neural network architectures as replacements for a Random Forest classifier in an operational context for the production of high resolution land cover maps with Sentinel-2 time-series at the country scale. Our contributions include a framework for working with Sentinel-2 L2A time-series image data, an adaptation of the U-Net model (a fully convolutional neural network) for dealing with sparse annotation data while maintaining high resolution output, and an analysis of those results in the context of operational production of land cover maps. We conclude that fully convolutional neural networks can yield improved results with respect to pixel-wise Random Forest classifiers for classes where texture and context are pertinent. However, this new approach shows higher variability in quality across different landscapes and comes with a computational cost which could be to high for operational systems.https://www.mdpi.com/2072-4292/11/17/1986land cover mappingconvolutional neural networksUNETSentinel-2 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Andrei Stoian Vincent Poulain Jordi Inglada Victor Poughon Dawa Derksen |
spellingShingle |
Andrei Stoian Vincent Poulain Jordi Inglada Victor Poughon Dawa Derksen Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems Remote Sensing land cover mapping convolutional neural networks UNET Sentinel-2 |
author_facet |
Andrei Stoian Vincent Poulain Jordi Inglada Victor Poughon Dawa Derksen |
author_sort |
Andrei Stoian |
title |
Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems |
title_short |
Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems |
title_full |
Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems |
title_fullStr |
Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems |
title_full_unstemmed |
Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems |
title_sort |
land cover maps production with high resolution satellite image time series and convolutional neural networks: adaptations and limits for operational systems |
publisher |
MDPI AG |
series |
Remote Sensing |
issn |
2072-4292 |
publishDate |
2019-08-01 |
description |
The Sentinel-2 satellite mission offers high resolution multispectral time-series image data, enabling the production of detailed land cover maps globally. When mapping large territories, the trade-off between processing time and result quality is a central design decision. Currently, this machine learning task is usually performed using pixel-wise classification methods. However, the radical shift of the computer vision field away from hand-engineered image features and towards more automation by representation learning comes with many promises, including higher quality results and less engineering effort. In particular, convolutional neural networks learn features which take into account the context of the pixels and, therefore, a better representation of the data can be obtained. In this paper, we assess fully convolutional neural network architectures as replacements for a Random Forest classifier in an operational context for the production of high resolution land cover maps with Sentinel-2 time-series at the country scale. Our contributions include a framework for working with Sentinel-2 L2A time-series image data, an adaptation of the U-Net model (a fully convolutional neural network) for dealing with sparse annotation data while maintaining high resolution output, and an analysis of those results in the context of operational production of land cover maps. We conclude that fully convolutional neural networks can yield improved results with respect to pixel-wise Random Forest classifiers for classes where texture and context are pertinent. However, this new approach shows higher variability in quality across different landscapes and comes with a computational cost which could be to high for operational systems. |
topic |
land cover mapping convolutional neural networks UNET Sentinel-2 |
url |
https://www.mdpi.com/2072-4292/11/17/1986 |
work_keys_str_mv |
AT andreistoian landcovermapsproductionwithhighresolutionsatelliteimagetimeseriesandconvolutionalneuralnetworksadaptationsandlimitsforoperationalsystems AT vincentpoulain landcovermapsproductionwithhighresolutionsatelliteimagetimeseriesandconvolutionalneuralnetworksadaptationsandlimitsforoperationalsystems AT jordiinglada landcovermapsproductionwithhighresolutionsatelliteimagetimeseriesandconvolutionalneuralnetworksadaptationsandlimitsforoperationalsystems AT victorpoughon landcovermapsproductionwithhighresolutionsatelliteimagetimeseriesandconvolutionalneuralnetworksadaptationsandlimitsforoperationalsystems AT dawaderksen landcovermapsproductionwithhighresolutionsatelliteimagetimeseriesandconvolutionalneuralnetworksadaptationsandlimitsforoperationalsystems |
_version_ |
1724728977728208896 |